JP2008211840A - Image processing apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processing apparatus capable of obtaining sufficient emphasis. <P>SOLUTION: An image processing apparatus includes: a non-sharp image signal creating section 10 for creating non-sharp image signals of a plurality of frequency bands from a source image signal; a differential processing section 12 for obtaining a differential image signal by making differential the source image signal or non-sharp image signal and a non-sharp image signal in a frequency band adjacent to these image signals; a density dependent transformation processing section 16 for applying transformation processing to the differential image signal so as to change a degree of processing while depending on a pixel value of the non-sharp image signal to be made differential; and an addition processing section 17 for adding the differential image signal to which the transformation processing has been applied by the density dependent transformation processing section, to the source image signal or to a bottom frequency image signal, or for adding a correction signal calculated from a low frequency component signal obtained by making differential a result integrating the image signal to which the trans formation processing has been applied by the density dependent transformation processing section, from the source image signal to the source image signal or to the bottom frequency image signal. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an image processing apparatus, and more particularly to a radiographic image processing apparatus.

  In the radiation image processing department, a method of applying a conversion process to an original image signal is used in order to obtain a better image. FIG. 15 is an explanatory diagram of a conventional frequency enhancement process. A non-sharp image signal 2 is generated from the original image signal 1, a difference image signal 3 is generated by subtracting the non-sharp image signal 2 from the original image signal 1, and the difference image signal 3 is multiplied by a predetermined coefficient β. A processed image signal is obtained by adding a signal as a high-frequency component signal to the original image signal 1.

  FIG. 16 is an explanatory diagram of a conventional dynamic range compression process. The same components as those in FIG. 15 are denoted by the same reference numerals. In this case, a difference image signal 3 is obtained by creating an unsharp image signal 2 from the original image signal 1 and subtracting the unsharp image signal 2 from the original image signal 1. On the other hand, density correction conversion 4 is added to the unsharp image signal 2 to obtain a corrected image signal 5. Then, the processed image signal is obtained by adding the difference image signal 3 to the obtained corrected image signal 5. In recent years, a method has been developed in which the above-described image processing method is improved to obtain a clearer processed image signal.

As a conventional apparatus of this type, a method of controlling the degree of addition in the high-frequency component signal of the original image low density part is used (for example, see Patent Document 1). Further, a method is used in which the enhancement degree of each band is determined based on the density of the image signal in the lowest frequency band that has been decomposed.
JP-A-10-75395 (paragraphs 0022 to 0023)

  The image signal is decomposed into a plurality of frequency bands, and each image signal in each band is converted and then added to the original image signal or the lowest frequency band image signal to enhance or convert the original image signal. The dynamic range of the image signal is compressed by adding the correction component signal created from the low frequency component signal created by subtracting the high frequency component signal created by adding the subsequent bands from the original image signal to the original image signal. An image signal processing method has been proposed.

  In these processes, edge enhancement and dynamic range correction are performed, but there is a problem in that the granularity of the low density portion is deteriorated by performing the process. Further, in the invention described in Patent Document 1, a method of controlling the degree of addition in the high frequency component signal of the low density part of the original image is used, but since each frequency band image does not include density information, The suppression is determined by the density of the original image signal, and all the high-frequency component signals are suppressed in the low density portion in order to perform the same suppression for all frequency bands.

  In addition, the method of determining the degree of emphasis of each band based on the density of the image signal of the lowest frequency band that has been decomposed is used. In this method, when the pyramid algorithm is used, the lowest image size that has the smallest image size is used. Since the size of the frequency band image signal is different from the size of the band image signal for suppressing the enhancement degree, there is a problem in that a separate correspondence has to be provided.

  The present invention has been made in view of such problems, and an object thereof is to provide an image processing apparatus capable of suppressing deterioration in granularity due to frequency processing and obtaining sufficient enhancement.

  (1) The invention described in claim 1 is a non-sharp image signal creating unit that creates unsharp image signals of a plurality of frequency bands from an original image signal composed of a plurality of pixels, and the original image signal or the unsharp image signal. A difference processing unit that obtains a difference image signal by subtracting the unsharp image signal created by the creation unit and the unsharp image signal in the frequency band adjacent to the original image signal or the unsharp image signal; and the difference processing unit A density-dependent conversion processing unit that performs conversion processing that changes the degree of processing on the obtained difference image signal depending on the pixel value of the unsharp image signal that is differenced by the difference processing unit; and the density-dependent conversion The difference image signal converted by the processing unit is added to the original image signal or the lowest frequency image signal, or the difference obtained by integrating the original image signal and the image signal converted by the density-dependent conversion processing unit is added. Frequency component signal The addition unit to obtain a processed image signal by adding to the original image signal or the lowest frequency image signal a correction signal calculated from and having a.

(2) The invention described in claim 2 is characterized in that the conversion processing unit suppresses an absolute value of a pixel value at least in part of an image signal.
(3) In the invention according to claim 3, the conversion processing unit performs a conversion process in which the absolute value of the image signal is more suppressed as the non-sharp image signal that is differenced by the difference processing unit is an image signal in a low frequency band. It is characterized by giving.

  (4) In the invention according to claim 4, the conversion processing unit performs a conversion process in which the absolute value of the image signal is largely suppressed as the non-sharp image signal that is differenced by the difference processing unit is a low-density image signal. It is characterized by giving.

  (1) According to the first aspect of the invention, the high-frequency component signal added to the original image signal is adjusted by performing density-dependent nonlinear conversion of the differential image signal, and noise and artifacts are suppressed together with edge enhancement. It is possible to create a processed image signal.

  (2) According to the invention described in claim 2, the emphasis in the high contrast portion which is the cause of the overshoot / undershoot is suppressed, and the edge emphasis and noise and artifacts can be more effectively suppressed.

  (3) According to the invention described in claim 3, in the differential image signal, the absolute value is more suppressed as the frequency band is lower, and the image signal is more effectively sharpened and noise and artifacts are suppressed. Is possible.

  (4) According to the invention described in claim 4, it is possible to suppress artifacts and noise with good sharpness by strongly suppressing the difference image signal as the density is lower.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing a first embodiment of the present invention for performing frequency enhancement processing. The apparatus shown in the figure obtains a processed image signal by adding a high-frequency component signal of an original image signal to a low-frequency signal of the original image signal or the original image signal for an original image signal composed of a plurality of pixels. 1 shows an image processing apparatus.

  In the figure, 10 is a non-sharp image signal creation unit that creates a non-sharp image signal from an original image signal, 11 is a conversion processing unit that performs a conversion process on the unsharp image signal created by the non-sharp image signal creation unit 10, and 12. Is a difference processing unit that takes a difference between the original image signal and the image signal after the conversion process and a difference between the unsharp image signal and the image signal after the conversion process, and 13 is a difference image signal obtained by the difference processing unit 12 Is an addition processing unit for adding. The unsharp image signal creation unit 10, the conversion processing unit 11, the difference processing unit 12, and the addition processing unit 13 can be realized by hardware or software. The operation of the apparatus configured as described above will be described as follows.

  The unsharp image signal creation unit 10 creates an unsharp image signal for the original image signal using, for example, a pyramid algorithm. The pyramid algorithm will be described later. From the non-sharp image signal creation unit 10, non-sharp image signals of a plurality of frequency bands having different frequency characteristics are obtained. The conversion processing unit 11 performs an image signal conversion process on the obtained unsharp image signal. Any known image signal conversion processing technique can be used as the image signal conversion processing.

  The difference processing unit 12 obtains the difference between the converted image signal and the original image signal thus obtained, and the non-sharp image signal and the converted image signal. Here, the obtained difference image signal is a difference between an unsharp image signal or an original image signal in an adjacent frequency band and an unsharp image signal after conversion. Next, the addition processing unit 13 adds the difference signals obtained by the difference processing unit 12 to obtain a high frequency component signal. A processed image signal is obtained by adding the high-frequency component signal thus obtained to the original image signal.

  Here, one specific process of the conversion processing unit 11 will be described. FIG. 2 is a diagram illustrating a correction function for an unsharp image. The horizontal axis x is the signal value representing the difference between the signal value before unsharpening and the signal value after unsharpening, the vertical axis y is the correction component signal, the upper is the low frequency, the lower Is a function representing the correction component signal in the high frequency band. This characteristic has a feature that many corrections are made in a region with a high contrast. The correction component signal thus obtained is added to the unsharp image signal to perform image conversion.

  This function varies depending on the density of the unsharp image signal as shown in FIG. In FIG. 3, the horizontal axis x is a signal value representing the difference between the signal value before unsharpening and the signal value after unsharpening, and the vertical axis y is a correction component signal. The upper part is a function representing the correction component signal in the low density, and the lower part is the correction component signal in the high density part. The conversion characteristics of the conversion processing unit 11 of this embodiment have both the characteristics shown in FIG. 2 and the characteristics shown in FIG. Therefore, by increasing the degree of correction in the lower density portion and the lower frequency band, the image signal approximates to the unsharp image signal or the original image signal one level higher.

  Thereby, the frequency emphasis in the low density part becomes weak, and the deterioration of the granularity can be suppressed. In addition, when the pixel difference from the image signal one level higher or the original image signal is large due to the averaging, a process for suppressing the averaging to bring the pixel closer to the image signal one level higher or the original image signal is added. Thus, overshoot / undershoot of the enhanced image signal can be suppressed. Further, this correction is strengthened as the frequency band becomes lower, so that a better image signal with good sharpness can be obtained.

  As described above, according to the configuration of FIG. 1, the difference image signal added to the original image signal is adjusted by converting the unsharp image signal, and the processed image signal in which noise and artifacts are suppressed together with edge enhancement is performed. Can be created.

  Further, the difference image signal is a difference between an unsharp image signal or an original image signal in an adjacent frequency band and an unsharp image signal after conversion, so that the frequency band of each difference image signal has less overlapping portions. By combining the conversion processing into a non-sharp image signal, operation for each band becomes possible.

  In addition, the conversion process performed on the non-sharp image signals in the plurality of frequency bands converts the pixel values of the non-sharp image signal based on a non-linear conversion, thereby effectively enhancing edge enhancement and suppressing noise and artifacts. It becomes possible.

  Further, the conversion process performed on the unsharp image signals in the plurality of frequency bands is determined by the original image signal or the unsharp image signals in the plurality of frequency bands, and thus depends on the tendency of unsharpening of the image signals. Processing can be performed, and edge enhancement and noise and artifacts can be suppressed more effectively.

  Further, the conversion process performed on the unsharp image signals in the plurality of frequency bands is determined by the unsharp image signal or the original image signal in the adjacent frequency band, and thus depends on the tendency of unsharpness of the image signal. Processing, and edge enhancement and noise and artifact suppression can be more effectively performed.

  In addition, the conversion process performed on the non-sharp image signals in the plurality of frequency bands varies depending on the pixel value of another unsharp image signal or original image signal used when creating the difference image signal. Thus, since the processing depending on the pixel before the unsharp processing is performed is possible, the conversion considering the high frequency component signal can be performed, and the edge enhancement and the suppression of the artifact and noise can be more effectively performed.

  Further, the conversion processing applied to the non-sharp image signals in the plurality of frequency bands is different depending on the unsharp image signal, so that adjustment depending on the frequency band is possible, and edge enhancement and noise and artifacts are more effectively performed. Can be suppressed.

  Further, the conversion process applied to the unsharp image signal is a conversion that suppresses the averaging of the image signal, thereby suppressing unsharpening in a high contrast portion that causes overshoot / undershoot, Edge enhancement and suppression of noise and artifacts can be performed more effectively.

  In addition, the conversion process applied to the unsharp image signal changes depending on the pixel value of the unsharp image signal, thereby enabling processing depending on the signal value of the unsharp image signal, and a signal (density that is easily noticeable). ) By enhancing the suppression of area artifacts, edge enhancement and noise and artifacts can be more effectively suppressed.

  In addition, the conversion process performed on the unsharp image signal changes depending on the pixel value of the unsharp image in the lowest frequency band, so that the change in conversion of the unsharp image signal follows the rough structure of the original image signal. It becomes possible.

  In the present invention, the conversion process performed on the non-sharp image signal changes according to the pixel value of the original image signal, so that the change of the unsharp image signal can faithfully follow the original image signal. In addition, the conversion process applied to the unsharp image signal increases the tendency of averaging suppression as the non-sharp image signal is in a lower frequency band, so that the averaging suppression is increased in a lower frequency band and sharper. It is possible to create a good image signal in which noise and artifacts are suppressed.

  In addition, the conversion process applied to the unsharp image signal increases the suppression of averaging as the non-sharp image signal is in a high frequency band, thereby averaging high frequency component signals that tend to contain more noise component signals. As a result, edge enhancement and noise and artifacts can be effectively suppressed.

  FIG. 4 is a block diagram showing a first embodiment of the present invention for performing dynamic range compression processing. This apparatus constitutes an image processing apparatus that obtains a processed image signal by adding a correction signal obtained from a low-frequency component signal of an original image signal to an original image signal or ultra-low-frequency image signal composed of a plurality of pixels. Yes. The same components as those in FIG. 1 are denoted by the same reference numerals.

  In the figure, 10 is a non-sharp image signal generating unit that generates a non-sharp image signal from an original image signal, 11 is a conversion processing unit that performs a conversion process on the image signal generated by the non-sharp image signal generating unit 10, and 12 is an original. A difference processing unit that takes a difference between the image signal and the image signal after the conversion process and a difference between the unsharp image signal and the image signal after the conversion process, and 14 integrates the difference signal obtained by the difference processing unit 12 A correction signal calculation unit that calculates a correction signal from a low-frequency component signal obtained by subtracting the high-frequency component signal obtained from the original image signal, and 15 performs a process of adding the correction signal obtained by the correction signal calculation unit It is a correction signal adding unit to be performed. The unsharp image signal creation unit 10, the conversion processing unit 11, the difference processing unit 12, the correction signal calculation unit 14, and the correction signal addition unit 15 can be realized by hardware or software. The operation of the apparatus configured as described above will be described as follows.

  The unsharp image signal creation unit 10 creates an unsharp image signal for the original image signal using, for example, a pyramid algorithm. From the non-sharp image signal creation unit 10, non-sharp image signals of a plurality of frequency bands having different frequency characteristics are obtained. The conversion processing unit 11 performs an image conversion process on the obtained unsharp image signal. Any known image conversion processing technique can be used as the image signal conversion processing.

  The difference processing unit 12 obtains the difference between the converted image signal and the original image signal thus obtained, and the non-sharp image signal and the converted image signal. Here, the obtained difference image signal is a difference between an unsharp image signal or an original image signal in an adjacent frequency band and an unsharp image signal after conversion. Next, the correction signal calculation unit 14 performs a process of calculating a correction signal on the difference signal obtained by the difference processing unit 12.

  Here, the correction component signal is determined as shown in FIG. 17, for example. FIG. 17 shows a correction component signal in the dynamic range compression of the low density portion. In the figure, the horizontal axis represents the signal value X of the unsharp image signal, and the vertical axis represents the correction component signal F (X). As shown in FIG. 17, a correction component signal that is larger as the signal value is lower is calculated and added to the original image signal. The correction signal addition unit 15 performs a process of adding the correction signal obtained by the correction signal calculation unit 14, and adds the correction signal to the original image signal to obtain a processed image signal.

  Also in this embodiment, each non-sharp image signal can be converted by a function as shown in FIG. The horizontal axis x is a signal value representing the difference between the signal value before unsharpening and the signal value after unsharpening, and the vertical axis y is a correction component signal. This function varies depending on the density of the unsharp image signal as shown in FIG. In FIG. 3, the horizontal axis x is a signal value representing a difference, and the vertical axis y is a correction component signal.

  In this way, by increasing the correction component signal in the lower density portion, the image signal is obtained by adding the high frequency component signal of the image signal. Thereby, the difference image signal does not include the high frequency component signal in the region corresponding to the low density. Therefore, the low frequency image signal obtained by subtracting the difference image signal from the original image signal includes a high frequency in the low density portion, and the high frequency component signal is also included when the dynamic range of the low frequency image signal is compressed. It is compressed together and granular deterioration due to dynamic range compression of the low density portion can be suppressed.

  In addition, when the pixel value of the image signal or the original image signal one level higher is obtained by averaging, a process for suppressing the averaging is performed so that the pixel approaches the image signal or the original image signal one level higher. Thus, it is possible to suppress overshoot / undershoot after processing. Further, a better image signal can be obtained by strengthening this correction as the low frequency band tends to have a larger pixel change due to averaging.

  The adjustment of averaging in the low density portion need not be performed for all unsharp image signals. By adjusting only the relatively high-frequency unsharp image signal, it is possible to suppress noise and adjust edge component signal enhancement. Further, the degree of adjustment can be changed by each non-sharp image signal. For example, it is also possible to create an image signal in which the edge component is preserved and the granularity is suppressed in the low-density portion by increasing the adjustment of the low-density portion as the non-sharp image signal including more high frequencies.

  According to this embodiment, the correction portion added to the original image signal or the ultra-low frequency image signal is adjusted by converting the unsharp image signal, and the dynamic range compression of the image signal and the suppression of noise and artifacts are adjusted. It is possible to create a processed image signal in which both are performed.

  Further, the difference image signal is a difference between an unsharp image signal or an original image signal in an adjacent frequency band and an unsharp image signal after conversion. The frequency band of the image signal has few overlapping portions, and the operation for each band can be performed by combining the conversion processing into the unsharp image signal.

  In addition, the conversion process performed on the non-sharp image signals in the plurality of frequency bands is to convert the pixel values of the non-sharp image signal based on the non-linear conversion. Noise and artifacts can be suppressed.

  Further, the conversion process applied to the unsharp image signals in the plurality of frequency bands is determined by the original image signal or the unsharp image signals in the plurality of frequency bands, thereby reducing the unsharpness of the image signal. Processing that depends on the trend can be performed, and noise and artifacts can be more effectively suppressed.

  Further, the conversion process performed on the unsharp image signals in the plurality of frequency bands is determined by the unsharp image signal or the original image signal in the adjacent frequency band, thereby reducing the unsharpness of the image signal. Processing that depends on the trend can be performed, and noise and artifacts can be suppressed more effectively.

  Further, the conversion process applied to the non-sharp image signals in the plurality of frequency bands varies depending on the pixel value of another unsharp image signal or original image signal used when the difference image signal is created. As a result, pixel-dependent processing before unsharp processing is possible, enabling conversion that takes higher frequency component signals into account, enabling more effective edge enhancement and suppression of artifacts and noise. It becomes.

  In addition, the conversion process performed on the non-sharp image signals in the plurality of frequency bands is different depending on the unsharp image signal, so that adjustment depending on the frequency band is possible, and noise and artifacts are more effectively performed. Can be suppressed.

  Further, the conversion process performed on the unsharp image signal is conversion that suppresses averaging of the image signal, thereby suppressing unsharpening in a high contrast portion that causes overshoot / undershoot, Noise and artifacts can be suppressed more effectively.

  In addition, since the conversion process performed on the unsharp image signal varies depending on the pixel value of the unsharp image signal, processing depending on the signal value of the unsharp image signal is possible, and noise is easily noticeable. Averaging can be suppressed in the signal region, and noise and artifacts can be suppressed more effectively.

  In addition, the conversion process applied to the unsharp image signal changes according to the pixel value of the unsharp image signal in the lowest frequency band, so that the change in conversion of the unsharp image signal follows the rough structure of the original image signal. Is possible.

Further, the conversion process applied to the unsharp image signal changes according to the pixel value of the original image signal, so that the change in conversion of the unsharp image can faithfully follow the original image signal.
In addition, the conversion process applied to the unsharp image signal increases the tendency of the suppression of averaging as the non-sharp image signal is in a lower frequency band, so that the suppression of averaging is increased in the lower frequency band. An image signal in which noise and artifacts with excellent sharpness are effectively suppressed can be created.

  In addition, the conversion process performed on the unsharp image signal is a high-frequency portion that tends to contain more noise component signals because the degree of suppression of averaging is stronger as the unsharp image signal is in a higher frequency band. As the suppression of averaging increases, noise and artifacts can be suppressed more effectively.

  According to the present invention, the non-sharp image signal density information is used to perform non-linear conversion on the difference between the original image signal and the unsharp image signal and the difference image signal that is the difference between the unsharp image signals. Is possible. FIG. 5 is a block diagram showing a second embodiment of the present invention for performing frequency enhancement processing. The same components as those in FIG. 1 are denoted by the same reference numerals. The apparatus shown in the figure obtains a processed image signal by adding a high-frequency component signal of the original image signal to the original image signal or a low-frequency signal of the original image signal for an original image signal composed of a plurality of pixels. An image processing apparatus is configured.

  In the figure, 10 is a non-sharp image signal generating unit that creates a non-sharp image signal from the original image signal, and 12 is a difference between the original image signal and the non-sharp image signal and a difference between the non-sharp image signal and the adjacent non-sharp image signal. A difference processing unit 16 takes a density-dependent conversion processing unit that performs a density-dependent conversion process on the difference signal obtained by the difference processing unit 12, and 17 denotes a converted image obtained by the density-dependent conversion processing unit 16. It is an addition processing unit that adds a signal to an original image signal. The non-sharp image signal creation unit 10, the difference processing unit 12, the density dependent conversion processing unit 16, and the addition processing unit 17 can be realized by hardware or software. The operation of the apparatus configured as described above will be described as follows.

  The unsharp image signal creation unit 10 creates an unsharp image signal for the original image signal using, for example, a pyramid algorithm. From the non-sharp image signal creation unit 10, non-sharp image signals of a plurality of frequency bands having different frequency characteristics are obtained. The difference processing unit 12 obtains the difference between the original image signal and the unsharp image signal thus obtained and the difference between the unsharp image signal and the adjacent unsharp image signal.

  Next, the density-dependent conversion processing unit 16 performs a density-dependent conversion process on the difference signal given from the difference processing unit 12. 6 and 7 are diagrams showing the conversion characteristics of the density-dependent conversion processing unit 16, FIG. 6 is a diagram showing a change in the frequency of the conversion function of the difference image signal, and FIG. 7 is a diagram showing a change due to the density.

  In FIG. 6, the horizontal axis x is a signal value representing a difference, and the vertical axis y is a pixel value of the converted differential image signal. In FIG. 7, the horizontal axis x is a signal value representing a difference, and the vertical axis y is a pixel value of the converted differential image signal. The density-dependent conversion processing unit 16 has a conversion function having both the characteristics shown in FIG. 6 and the characteristics shown in FIG. FIG. 6 is a function in which the upper characteristic represents conversion in a high frequency and the lower characteristic represents conversion in a low frequency band. FIG. 7 shows a case where the upper characteristic is a high density when the non-sharp image used when creating the difference image signal, and the lower characteristic is when the non-sharp image used when creating the difference image signal is a low density. This is a function that represents conversion.

  In this case, since the image sizes of the difference image signal and the unsharp image signal match, it is easy to obtain the pixel value of the unsharp image signal corresponding to the pixel of the difference image signal. In this case, the non-linear function can suppress artifacts such as overshoot and undershoot by suppressing the signal where the difference value is large as shown in FIGS. 6 and 7. Also, by suppressing the difference component signal more strongly in the lower frequency band and lower density portion, it is possible to obtain a better image signal with better sharpness and less artifacts and noise.

  Next, the addition processing unit 17 performs addition processing on the signals thus converted. Then, this addition signal is added to the original image signal. As a result, it is possible to create a processed image signal in which the high-frequency component signal added to the original image signal is adjusted by performing density-dependent nonlinear conversion of the difference image signal, and noise and artifacts are suppressed together with edge enhancement. It becomes.

  FIG. 8 is a block diagram showing a second embodiment of the present invention for performing dynamic range compression processing. The same components as those in FIGS. 4 and 5 are denoted by the same reference numerals. The apparatus shown in the figure is an image processing apparatus that obtains a processed image signal by adding a correction signal obtained from a low-frequency component signal of the original image signal to an original image signal or ultra-low-frequency image signal composed of a plurality of pixels. It is composed.

  In the figure, 10 is a non-sharp image signal creation unit that creates a non-sharp image signal from the original image signal, and 12 is a difference between the original image signal and the non-sharp image signal and a difference between the non-sharp image signal and the adjacent unsharp image signal. A difference processing unit 16 that performs a density-dependent conversion process on the difference signal obtained by the difference processing unit 12, and 14 indicates a conversion obtained by the density-dependent conversion processing unit 16. A correction signal calculation unit for calculating a correction signal from a low frequency component signal obtained by subtracting a high frequency component signal obtained by integrating the image signals from the original image signal, and 15 indicates a correction signal obtained by the correction signal calculation unit. It is a correction signal adding unit for adding. The non-sharp image signal creation unit 10, the difference processing unit 12, the density-dependent conversion processing unit 16, the correction signal calculation unit 14, and the correction signal addition unit 15 can be realized by hardware or software. The operation of the apparatus configured as described above will be described as follows.

  The unsharp image signal creation unit 10 creates an unsharp image signal for the original image signal using, for example, a pyramid algorithm. From the non-sharp image signal creation unit 10, non-sharp image signals of a plurality of frequency bands having different frequency characteristics are obtained. The difference processing unit 12 obtains a difference between the converted image signal thus obtained, the original image signal, and the unsharp image signal adjacent to the unsharp image signal.

  Next, the density-dependent conversion processing unit 16 performs a density-dependent conversion process on the difference signal given from the difference processing unit 12. As the conversion characteristics at this time, the conversion characteristics shown in FIGS. 6 and 7 are used. The correction signal calculation unit 14 calculates a correction signal from a low frequency component signal obtained by subtracting a high frequency component signal obtained by integrating the converted image signals obtained by the density dependent conversion processing unit 16 from the original image signal. The correction signal adding unit 15 adds the correction signal thus obtained to the original image signal. Then, a processed image signal is obtained by adding this correction signal to the original image signal.

  According to this embodiment, the high-frequency component signal added to the correction component signal added to the original image signal or the ultra-low-frequency image signal is adjusted by converting the differential image signal and suppressing the component signal. Thus, it is possible to obtain a processed image signal in which both compression of the dynamic range of the image signal and suppression of noise and artifacts are performed.

  Further, the difference image signal is a difference between an unsharp image signal or an original image signal in an adjacent frequency band and an unsharp image signal after conversion. The frequency bands of the signals are less overlapped with each other, and the operation for each band can be performed by combining the conversion processing into an unsharp image signal.

  In addition, the non-sharp image signal on which the conversion process depends is an image signal used when obtaining the difference image signal. Therefore, even when the pyramid algorithm is used, the non-sharp image signal having the same image size as the converted image signal is used. A sharp image signal can be used, and the processing can be simplified.

  In addition, since the conversion processing applied to the plurality of difference image signals differs depending on the difference image signal, adjustment depending on the frequency band is possible, and edge enhancement and noise and artifact suppression are more effectively performed. It becomes possible.

  Further, the conversion process performed on the difference image signal suppresses the absolute value of the pixel value in at least a part of the image signal, thereby enhancing the high contrast portion that causes overshoot / undershoot. Is suppressed, and edge enhancement and noise and artifacts can be suppressed more effectively.

  Further, the conversion process applied to the difference image signal is such that the lower the frequency band of the difference image signal, the greater the suppression of the absolute value of the image signal. As the absolute value is suppressed, the better the image signal, the better the sharpness and the fewer artifacts and noise.

  The conversion process performed on the difference image signal is a high-frequency portion that tends to contain more noise component signals because the absolute value of the image signal is more suppressed as the difference image signal is in a higher frequency band. As the absolute value is suppressed, edge enhancement and noise and artifacts can be suppressed more effectively. By the way, the high frequency component signal in the present invention is a signal obtained by integrating the difference image signals.

  Next, the function of the user interface given to the user will be described. As a frequency characteristic designation method according to the present invention, for example, the one shown in FIG. 9 is conceivable. In the figure, the horizontal axis represents the frequency band, and the vertical axis represents the enhancement degree. As shown in FIG. 9, the emphasis degree of each frequency band is designated by a to f in the figure. This can be specified with a mouse or the like, or a numerical value may be specified. The term “adjacent” in the above claims refers to images that are adjacent when arranged in descending order of frequency band including unsharp image signals or differential image signals. The frequency components included in the difference image signal are not necessarily completely separated, and there may be cases where component signals included in each other overlap. In addition, there may be a case where some frequency bands are not included and are adjacent to each other with a gap.

  Further, the degree of emphasis on density can be specified by a graph as shown in FIG. In the figure, the horizontal axis represents density, and the vertical axis represents enhancement degree. The frequency characteristic and the emphasis level by density can be set for each band, or the overall frequency characteristic is determined and the relationship between the density and emphasis level common to all frequency bands to be emphasized is set. It is good. The graph can be specified by specifying the degree of enhancement in A and B, for example.

  In this way, by designating the frequency characteristics, a conversion function that realizes the given frequency characteristics is determined, and by performing processing using the determined conversion function, the user can set various parameters necessary for setting. It is only necessary to specify a desired frequency characteristic without being aware of the above, and the processing can be simplified.

  In addition, since the specification of the frequency characteristic can be changed depending on the density, the user operates the frequency characteristic of the image signal area corresponding to the density where noise is conspicuous, and suppresses noise enhancement. The specified process can be specified more easily.

  In addition, the frequency characteristics can be changed depending on the density for each non-sharp image signal or differential image signal, so that the user can easily set the processing intensity depending on the image signal value corresponding to the density for each frequency band. it can.

  In addition, it possesses a set of parameters necessary for the frequency characteristic processing, and by specifying the processing by selecting the parameter set, the user can easily handle the most suitable parameter set without handling many parameters. Can be selected.

  Next, the pyramid algorithm will be described. FIG. 11 is a block diagram illustrating a configuration example of a decomposition unit that executes the pyramid algorithm. In the figure, symbol ↑ indicates interpolation processing, symbol ↓ indicates downsampling, and F indicates filter processing. Here, a case where differential image signals b0 to bL-1 are obtained will be described.

  As shown in the figure, when the digital image signal S representing the original image signal is input to the multi-resolution decomposition processing means 30, the filtering means 20 filters it with a low-pass filter. The low-pass filter substantially corresponds to a two-dimensional Gaussian distribution on a 5 × 5 grid as shown in FIG. 12, for example. The image signal S filtered by such a low-pass filter is sampled every other pixel in the filtering means 20, and a low resolution approximate image signal g1 is obtained.

  The low-resolution approximate image signal g1 is ¼ the size of the original image signal. Subsequently, the interpolation unit 21 interpolates pixels having a value of 0 at the sampled interval of the low resolution approximate image signal g1. This interpolation is performed by inserting a row and a column having a value of 0 for every column and every row of the low resolution approximate image signal g1. In this way, the low-resolution approximate image signal g1 interpolated with the pixel having the value of 0 is blurred, but since the pixel having the value of 0 is inserted every other pixel, the change in the signal value is not smooth. It has become.

  Then, after the interpolation is performed in this way, the low resolution approximate image signal g1 subjected to the interpolation is again subjected to the filtering process by the low pass filter shown in FIG. 12 to obtain the low resolution image signal g1 '. The low-resolution approximate image signal g1 'has a smooth signal value change as compared with the low-resolution approximate image signal g1 subjected to the above-described interpolation.

  Also, the image signal is such that a frequency higher than half has disappeared in terms of the frequency band compared to the original image signal. This is because the image size is reduced to ¼, pixels with a value of 0 are interpolated every other pixel, and filtering processing is performed by the low-pass filter shown in FIG. This is because the image signal in the high frequency band is blurred.

  Next, the subtracter 22 subtracts the low resolution approximate image signal g1 'from the original image signal to obtain a difference image signal b0. This subtraction is performed between signals for pixels corresponding to the original image signal and the low-resolution approximate image signal g1 '. Here, since the low-resolution approximate image signal g1 ′ is such that an image in a frequency band higher than half of the spatial frequency of the original image signal is blurred as described above, the difference image signal b0 is included in the original image signal. The image signal represents only the frequency band above half. That is, as shown in FIG. 13, the differential image signal b0 represents an image signal in the frequency band N / 2 to N of the Nyquist frequency N of the original image signal.

  Next, the low resolution approximate image signal g1 is input to the filtering means 20 and subjected to filtering processing by a low pass filter shown in FIG. The filtered low resolution approximate image signal g1 is sampled every other pixel by the filtering means 20, and a low resolution approximate image signal g2 is obtained. The low-resolution approximate image signal g2 is 1/4 the size of the low-resolution approximate image signal g1, that is, 1/16 of the original image signal.

  Next, the interpolation unit 21 interpolates pixels having a value of 0 at the sampled interval of the low resolution approximate image signal g2. This interpolation is performed by inserting rows and columns having a value of 0 for each column and every row of the low resolution approximate image signal g2. As described above, the low-resolution approximate image signal g2 obtained by interpolating the pixels having the value 0 is blurred, but the pixels having the value 0 are inserted every other pixel, so that the change in the signal value is not smooth. It has become.

  Then, after the interpolation is performed in this way, the low resolution approximate image signal g2 subjected to the interpolation is further subjected to the filtering process by the low pass filter shown in FIG. 12 to obtain the low resolution approximate image signal g2 ′. . The low-resolution approximate image signal g2 'has a smooth signal value change as compared with the low-resolution approximate image signal g2 subjected to the above-described interpolation. In addition, the image signal in the frequency band higher than half of the frequency band in comparison with the low resolution approximate image signal g1 disappears.

  Next, the subtracter 22 subtracts the low resolution approximate image signal g2 'from the low resolution approximate image signal g1 to obtain a difference image signal b1. This subtraction is performed between the signals for the corresponding pixels of the low resolution approximate image signal g1 and the low resolution approximate image signal g2 '. Here, since the low resolution approximate image signal g2 ′ is such that an image in a frequency band higher than half of the spatial frequency of the low resolution approximate image signal g1 is blurred as described above, the difference image signal b1 is low. It is an image signal representing only a frequency band above half of the resolution approximate image signal g1.

  That is, as shown in FIG. 13, the difference image signal b1 is only in the frequency band above half of the low resolution approximate image signal g1, that is, the frequency N / 4 to N / 2 of the Nyquist frequency N of the original image signal. It represents a band image signal. In this way, the difference image signal is obtained by performing the filtering process by the low-pass filter of the Gaussian distribution. However, since the image signal subjected to the filtering process is subtracted from the low-resolution approximate image signal, The result is the same as when filtering is performed using a Laplacian filter.

Then, the above-described processing is sequentially repeated on the low resolution approximate image signal gk (k = 0 to L-1) filtered and sampled by the filtering means 20, and n differential image signals are obtained as shown in FIG. bk (k = 0 to L-1) and the residual image signal gL of the low resolution approximate image signal are obtained. Here, the resolution of the difference image signal bk decreases in order from b0. That is, the frequency band of the image signal is lowered, and the differential image signal bk represents the frequency band of N / 2 ^{k + 1 to} N / 2 ^k with respect to the Nyquist frequency N of the original image signal, and the size of the image Is 1/2 ^2k times the original image.

  That is, the difference image signal b0 having the highest resolution is the same size as the original image signal, but the difference image signal b1 having the next highest resolution after the difference image signal b0 is 1/4 of the original image signal. As described above, since the difference image signal sequentially decreases from the same size as the original image signal, and the difference image signal is substantially the same image signal as that subjected to the Laplacian filter. The multi-resolution conversion according to the embodiment is also called a Laplacian pyramid algorithm.

  Further, the residual image signal gL can be regarded as an approximate image signal with a very low resolution of the original image signal. In an extreme case, the residual image signal gL (corresponding to the above-mentioned lowest frequency image signal, the original image signal The image signal obtained as a result of the final filtering process is executed from a single image signal representing the average value of the original image signal. Will be. The difference image signal bk obtained in this way is stored in a memory (not shown). The above-described image signal conversion processing according to the present invention is performed on g1 ', g2', g3 ',... Output from the interpolation means 21 shown in FIG. Alternatively, it is performed for b0, b1, b2,. These unsharp image signals g1 ', g2', g3 '... are unsharp image signals of a plurality of frequency bands having different frequency characteristics. In the present invention, the image processing as described above is performed using these unsharp image signals.

  Further, instead of using such 0 interpolation and low-pass filter, interpolation by the following method may be used. First, interpolation processing by weighting according to linear interpolation, spline interpolation, or cardinal sign (sampling function) may be performed on the columns, and then the same processing may be performed on the rows.

The difference image signal bk in the frequency band on which the image conversion processing according to the present invention has been performed and the difference image signal in another frequency band are inversely transformed. Here, a case where signals b0 to bL-1 are input is shown.
This inverse conversion process is performed by the restoration processing means 40. FIG. 14 is a block diagram illustrating a configuration example of a restoration unit that executes the pyramid algorithm. First, the image signal bL-1 is interpolated between the pixels by the interpolation means 24 to obtain an image signal bL-1 'having a size four times the original size. Next, the adder 25 adds the interpolated image signal bL-1 ′ and the corresponding pixel of the lowest resolution difference image signal bL-2, and adds the added image signal (bL-1 ′ + bL-2). )

  Next, the added image signal (bL-1 ′ + bL-2) is input to the interpolation unit 24, and the interpolation unit 24 interpolates between the pixels to obtain an image signal bL-2 that is four times the original size. It is said. Next, the image signal bL-2 ′ is subjected to addition of pixels corresponding to the difference image signal bL-3 of the one-step high resolution of the difference image signal bL-2 in the adder 25, and the added addition signal is added. (BL−2 ′ + bL−3) is interpolated by the interpolating unit 24 to obtain an image signal bL−3 ′ that is four times larger than the difference image signal bL−3.

  Thereafter, the same processing is repeated. Then, this process is sequentially performed on the higher-frequency difference image signal, and finally, the adder 25 adds the interpolated image signal b1 ′ and the highest-resolution difference image signal b0 to β times by the multiplier 26, The adder 29 adds the original image signal S to the processed image signal signal Sout (frequency enhancement process). Alternatively, density correction is performed on the original image signal S obtained by subtracting b0 ', and the result is added to the original image signal S by the adder 29 to obtain a processed image Sout (dynamic range compression processing).

  The processed image signal signal S ′ thus obtained is input to the image signal output means and displayed as a visible image. The image signal output means may be a display means such as a CRT or a recording apparatus that performs optical scanning recording on a photosensitive film.

  In the above-described embodiment, the pyramid algorithm is used as the decomposition method into the multi-resolution space, but the present invention is not limited to this. For example, a scaling function or wavelet transform / inverse transform may be used. When wavelet transform is used, for example, it is possible to perform enhancement processing in an arbitrary direction (vertical direction, horizontal direction, diagonal direction).

  In the above-described embodiment, the conversion function differs depending on the density of the unsharp image signal to be converted. However, this conversion function may be the unsharp image signal or the original image signal in the lowest frequency band.

  As described above, according to the present invention, it is possible to provide an image processing apparatus that can suppress deterioration in granularity due to frequency processing and obtain sufficient enhancement.

It is a block diagram which shows the 1st Example of this invention which performs a frequency emphasis process. It is a figure which shows the correction function of a non-sharp image signal. It is a figure which shows a density | concentration and a correction function. It is a block diagram which shows the 1st Example of this invention which performs a dynamic range compression process. It is a block diagram which shows the 2nd Example of this invention which performs a frequency emphasis process. It is a figure showing the change regarding the frequency of the conversion function of a difference image signal. It is a figure showing the change by a density | concentration. It is a block diagram which shows the 2nd Example of this invention which performs a dynamic range compression process. It is explanatory drawing of designation | designated of a frequency characteristic. It is explanatory drawing of designation | designated of density dependence emphasis. It is a block diagram which shows the structural example of the decomposition | disassembly part which performs a pyramid algorithm. It is a figure showing a low-pass filter. It is explanatory drawing of the magnitude | size of the output image signal of a pyramid algorithm decomposition | disassembly part. It is a block diagram which shows the structural example of the decompression | restoration part which performs a pyramid algorithm. It is explanatory drawing of the conventional frequency emphasis process. It is explanatory drawing of the conventional dynamic range compression process. It is explanatory drawing of correction | amendment component signal calculation.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Non-sharp image signal preparation part 11 Conversion processing part 12 Difference processing part 13 Addition processing part

Claims (4)

A non-sharp image signal creating unit that creates a non-sharp image signal of a plurality of frequency bands from an original image signal composed of a plurality of pixels;
The difference between the original image signal or the unsharp image signal created by the unsharp image signal creation unit and the unsharp image signal in the frequency band adjacent to the original image signal or the unsharp image signal is obtained. A difference processing unit;
A density-dependent conversion processing unit that performs a conversion process in which the degree of processing changes depending on the pixel value of the non-sharp image signal that is differenced by the difference processing unit with respect to the difference image signal obtained by the difference processing unit; ,
The difference image signal converted by the density-dependent conversion processor is added to the original image signal or the lowest frequency image signal, or the image signal converted by the density-dependent conversion processor is integrated from the original image signal An addition processing unit that obtains a processed image signal by adding a correction signal calculated from a low-frequency component signal obtained by subtracting the original image signal or the lowest frequency image signal;
An image processing apparatus comprising:   The image processing apparatus according to claim 1, wherein the conversion processing unit suppresses an absolute value of a pixel value in at least a part of the image signal.   The conversion processing unit performs a conversion process in which the absolute value of the image signal is more greatly suppressed as the unsharp image signal differenced by the difference processing unit is an image signal in a low frequency band. Image processing apparatus.   4. The conversion processing unit according to claim 2 or 3, wherein the conversion processing unit performs conversion processing in which the absolute value of the image signal is largely suppressed as the non-sharp image signal differenced by the difference processing unit is a low-density image signal. The image processing apparatus described.

Images