JP2011009857A - Noise level measuring apparatus and image processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately measure random noise included in stereo images.SOLUTION: A measurement pertinence determination part 3 divides respective parallax images (left image SL, right image SR) into a plurality of small blocks and determines whether or not each of the plurality of small blocks is suitable for noise level measurement. A differential variance value calculation part 2 obtains a differential variance value between the two parallax images by small block unit. A histogram generation part 4 generates the histogram of the differential variance value by adding only the values for the small blocks determined to be suitable for the noise level measurement among the differential variance values obtained by the differential variance value calculation part. A noise level estimation part 5 analyzes the histogram and estimates the noise level of the random noise included in common in the two parallax images.

Description

  The present invention relates to a noise level measurement device and an image processing device having a function of measuring a noise component included in a parallax image.

  Conventionally, an apparatus for measuring a noise component included in an input video signal and removing the noise based on the measurement result has been developed. Patent Document 1 proposes a method of generating a histogram using a temporal frame difference image and estimating a noise level by histogram analysis. Patent Document 2 proposes a method of determining a region suitable for noise level measurement based on an edge amount, a motion amount, or the like between frames in a time direction, and limiting a region for performing histogram addition.

JP 2008-294696A JP 2009-3599 A

  By the way, in recent years, two cameras are used for security use or in-vehicle use, and two parallax images (two left and right stereo images) taken at the same time are obtained from each camera, and a stereo image is obtained by a stereo method. The number of devices that perform recognition is increasing. In such an apparatus, depth information can be acquired by using two parallax images, so that a person or an object can be recognized three-dimensionally. It is considered that security cameras and in-vehicle cameras used in such devices are often photographed in situations where photographing conditions are poor and noise components are likely to be mixed, such as photographing in a dark place. Therefore, in order to perform image recognition correctly, it is necessary to correctly remove noise components from the captured stereo image.

  On the other hand, the conventional noise level measurement methods described in Patent Documents 1 and 2 are methods for estimating the noise level between frames in the time direction. For this reason, when applied to a stereo image, for example, the noise level is measured in the time direction for each of the left image and the right image. In this case, a frame memory is required for each of the left image and the right image, and a frame delay occurs in reflection when the intensity of noise reduction or the like is controlled using the obtained noise level. In addition, in a stereo image on which a difference is assumed to occur, the difference value of the signal component itself may be erroneously recognized as a noise component unless the measurement region is limited in a form corresponding to the feature of the stereo image. .

  The present invention has been made in view of such problems, and an object of the present invention is to provide a noise level measurement apparatus capable of accurately measuring random noise included in a stereo image, and a stereo image based on the measurement result. It is another object of the present invention to provide an image processing apparatus that can satisfactorily remove random noise components.

  The noise level measurement apparatus according to the present invention receives two parallax images with parallax from each other, divides each parallax image into a plurality of small blocks, and whether each of the plurality of small blocks is suitable for noise level measurement. A measurement appropriateness determination unit, a difference dispersion value calculation unit for obtaining a difference variance value between two parallax images in units of small blocks, and a difference appropriateness value among the difference variance values obtained by the difference variance value calculation unit Two disparities by analyzing the histogram generated by the histogram generation unit, which generates a histogram of difference variance values by adding only values for small blocks determined to be suitable for noise level measurement in the unit A noise level estimation unit that estimates a noise level of random noise included in common in the image.

  An image processing apparatus according to the present invention includes a noise level measurement unit that measures a noise level of random noise that is included in common between two parallax images having parallax, and two parallax images based on the measured noise level. A noise removing unit that removes random noise is provided for each, and the noise level measuring device of the present invention is applied to the noise level measuring unit.

  In the noise level measurement apparatus or the image processing apparatus according to the present invention, each parallax image is divided into a plurality of small blocks, and it is determined whether or not each of the plurality of small blocks is suitable for noise level measurement. In addition, a difference variance value between two parallax images is obtained in units of small blocks, and only a value for a small block determined to be suitable for noise level measurement is added to generate a histogram of the difference variance value. Based on the histogram, the noise level of random noise included in common to the two parallax images is estimated.

  According to the noise level measuring device or the image processing device of the present invention, the difference variance value is obtained by adding only the value for the small block determined to be suitable for noise level measurement to the difference variance value between two parallax images. Since the noise level of random noise is estimated based on the histogram, the random noise included in the stereo image can be measured with high accuracy. In particular, since the measurement area of the noise level is limited according to the feature of the stereo image, it is possible to prevent the difference value of the signal component itself that is not noise from being erroneously recognized as the noise component.

  In particular, according to the image processing apparatus of the present invention, random noise is removed based on the measurement result of the noise level measurement apparatus of the present invention, so that random noise can be favorably removed from a stereo image. As a result, for example, when performing stereo image recognition by the stereo method, it is possible to suppress deterioration in recognition performance of an object even in a stereo image obtained under a bad shooting condition such as a security camera or an in-vehicle camera. .

It is a block diagram which shows the structural example of the noise level measuring apparatus which concerns on the 1st Embodiment of this invention. It is a block diagram which shows the structural example of the difference dispersion value calculation part in the noise level measurement apparatus which concerns on 1st Embodiment. It is a block diagram which shows the structural example of the measurement appropriateness determination part in the noise level measuring device which concerns on 1st Embodiment. It is a block diagram which shows the structural example of the noise level estimation part in the noise level measuring device which concerns on 1st Embodiment. It is explanatory drawing which shows notionally the technique of the noise level measurement by the noise level measuring apparatus which concerns on 1st Embodiment. It is explanatory drawing which shows notionally the method of the appropriateness determination by a measurement appropriateness determination part. It is explanatory drawing which shows notionally the method of the appropriateness determination using a block average. It is explanatory drawing which shows notionally the method of the appropriateness determination using edge detection. It is a block diagram which shows the structural example of the image processing apparatus which concerns on the 2nd Embodiment of this invention.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

<First Embodiment>
[Configuration of noise level measurement device]
FIG. 1 shows a configuration example of a noise level measuring apparatus 1 according to the first embodiment of the present invention. FIG. 5 conceptually shows a noise level measurement method by the noise level measurement apparatus 1. This noise level measuring apparatus 1 is a random noise in two parallax images (two left and right stereo images (left image SL, right image SR)) having parallax taken at the same time by, for example, two left and right cameras. The noise level is measured. In the noise level measuring apparatus 1, the noise level is estimated using a difference image between the left image SL and the right image SR. This is because, for example, when shooting with two cameras on the left and right, the two cameras are shooting under almost the same conditions, so the noise levels included in the stereo images obtained from the two cameras are almost the same. It is based on the idea that

  The noise level measurement device 1 includes a difference variance value calculation unit 2, a measurement appropriateness determination unit 3, a histogram generation unit 4, and a noise level estimation unit 5. Signals indicating two parallax images (the left image SL and the right image SR) are input to the difference variance value calculation unit 2 and the measurement appropriateness determination unit 3.

The difference variance value calculation unit 2 calculates a difference variance value between two parallax images in units of small blocks 31. As shown in FIGS. 2 and 5, the difference variance value calculation unit 2 calculates a difference unit 6 that obtains a difference image between the left image SL and the right image SR, and an intra-block variance value (vibration component) in the difference image. And an intra-block variance calculating unit 7. The dispersion value here means an average absolute deviation MAD, a standard dispersion value σ ² and the like as described below. dy is a luminance difference, ave is an average in the block with respect to the luminance difference value, and N is the number of pixels in the block.
MAD = Σ | dy-ave | / N
σ ² = Σ {dy-ave} ² / N

  The measurement appropriateness determination unit 3 divides the left image SL and the right image SR into the same number of small blocks 31L and 31R of M in the vertical direction and N in the horizontal direction, and each of the small blocks 31L and 31R. It is judged whether it is suitable for noise level measurement. For example, as shown in FIG. 6, the difference image between the left image SL including the object 32 </ b> L and the right image SR including the object 32 </ b> R can be roughly divided into an object portion 32 and a background image portion 33. When measuring the noise level of such a difference image, the difference between the signal components tends to be large in the vicinity of the object portion 32, and the signal component may be erroneously measured as a noise component. It is preferable to exclude. Conversely, a portion where the difference between the signal components is small, such as the background image portion 33, can be determined to be suitable for noise level measurement. The measurement appropriateness determination unit 3 determines whether or not it is suitable for noise level measurement based on such an idea.

  FIG. 3 shows a configuration example of the measurement appropriateness determining unit 3. The measurement appropriateness determination unit 3 includes an intra-block variance calculation unit 7, a first intra-block average calculation unit 8, a second intra-block average calculation unit 9, a first edge detection unit 10, and a second edge It has a detection unit 11, a stereo matching circuit 12, and a determination unit 13.

  The first intra-block average calculator 8 calculates the intra-block average of the luminance in units of small blocks 31L for the left image SL. The second intra-block average calculation unit 9 calculates an intra-block average of luminance for each small block 31R for the right image SR. As shown in FIG. 7, the determination unit 13 excludes the noise level measurement when the luminance average values of the small blocks 31L and 31R at the corresponding positions in the left image SL and the right image SR are greatly different. It is determined that it is not suitable for noise level measurement. This is because the random noise component itself has an average value of 0 due to its randomness, and the difference in luminance average value between the two images is considered to be a difference in the signal component itself. It is. The specific determination process in the determination unit 13 in this case is performed by determining whether or not the difference value of the average value in the block between the left and right images is larger than a predetermined threshold value. When it is larger than the predetermined threshold, it is determined that it is not suitable for noise level measurement, and when it is smaller, it is determined that it is suitable for measurement.

  The first edge detection unit 10 performs edge detection on the left image SL in units of small blocks 31L. The second edge detection unit 11 performs edge detection on the right image SR in units of small blocks 31R. As shown in FIG. 8, when the determination unit 13 includes an edge component in any one of the small blocks 31L and 31R at the corresponding positions in the left image SL and the right image SR, It is determined that the block is not suitable for noise level measurement. This is because, when the difference is taken between the left and right images, a difference is particularly likely to occur near the edge of the object. Conventionally, as an edge detection method, there are a method of determining based on the magnitude of the luminance difference between adjacent pixels, a method of determining based on the size of the high-frequency extraction filter, and the variance within the block, and any method may be used.

  The stereo matching circuit 12 performs image analysis by the stereo matching method for each of the small blocks 31 for the left image SL and the right image SR. The determination unit 13 determines the likelihood of the background in units of small blocks 31 based on the image analysis result by the stereo matching method, and determines that the small block 31 corresponding to the background-like portion is suitable for noise level measurement. Conversely, when searching for the corresponding points between the left and right images using the stereo matching method, if an analysis result is obtained in which the coordinate positions of the corresponding points are greatly separated between the two images, Is low and is estimated to be the object portion 32 as shown in FIG. The determination unit 13 determines that such a small block 31 is not suitable for noise level measurement.

  The determination unit 13 finally adds up the results of the determination of the appropriateness of measurement using the above average in block, edge detection, and stereo matching. Then, a histogram addition control signal for excluding measurement of small blocks not suitable for measurement is generated and output to the histogram generation unit 4. The determination unit 13 further counts the number of blocks excluded from measurement in the screen, and outputs a measurement suitability determination result including the count result to the noise level estimation unit.

  The histogram generation unit 4 generates a histogram of difference variance values in units of small blocks 31 in the difference image between the left image SL and the right image SR obtained by the difference variance value calculation unit 2. As shown in FIG. 5, the histogram generation unit 4 generates a histogram by adding only the values for the small blocks 31 determined to be suitable for noise level measurement by the measurement appropriateness determination unit 3 among the difference variance values. It is like that. As a result, only the values for the small blocks 31 determined to be suitable for noise level measurement are added as a histogram, so that the parameter added to the histogram is reduced, but represents a more accurate distribution of noise components. A histogram will be obtained.

  The noise level estimation unit 5 analyzes the histogram generated by the histogram generation unit 4 and estimates the noise level of random noise included in both the left image SL and the right image SR. As illustrated in FIG. 4, the noise level estimation unit 5 includes a peak determination unit 14, a reliability calculation unit 15, and a multiplier 16. The peak value generated by the histogram generation unit 4 is input to the peak determination unit 14 and the reliability calculation unit 15. The reliability calculation unit 15 is also input with data of a measurement suitability determination result including information on the number of small blocks from which the noise level measurement is excluded, which is output from the measurement suitability determination unit 3.

  The peak determination unit 14 estimates the noise level of random noise based on the input histogram. The peak determination unit 14 will be described later. For example, the noise level is estimated based on the variance value having the highest frequency in the histogram or the variance value serving as the center of gravity in the histogram. For example, the reliability calculation unit 15 calculates the reliability of the noise level estimated by the peak determination unit 14 from the likelihood of normal distribution of a histogram. The reliability calculation unit 15 is also estimated by the peak determination unit 14 based on the information (the number of small blocks excluded from measurement) of the small blocks 31 determined not suitable for noise level measurement by the measurement appropriateness determination unit 3. The reliability of the noise level is calculated.

  The noise level estimator 5 corrects the value of the noise level estimated by the peak determiner 14 according to the reliability of the noise level calculated by the reliability calculator 15. Specifically, the multiplier 16 corrects the noise level value by multiplying the noise level output from the peak determination unit 14 by the reliability of the noise level. The noise level estimation unit 5 outputs the output from the multiplier 16 as the final value of the noise level. The noise level estimator 5 also outputs the reliability of the noise level together with the final noise level.

[Operation of Noise Level Measuring Device 1]
In the noise level measuring apparatus 1, the measurement appropriateness determining unit 3 divides the left image SL and the right image SR into a plurality of small blocks 31L and 31R, and each of the plurality of small blocks 31L and 31R is suitable for noise level measurement. It is determined whether or not. Further, the difference variance value calculation unit 2 obtains a difference variance value for each small block 31 in the difference image between the left image SL and the right image SR. The difference variance value is added only to the small block 31 determined to be suitable for noise level measurement by the histogram generation unit 4, thereby generating a histogram of the difference variance value. The noise level estimation unit 5 estimates the noise level of random noise included in common in the two parallax images based on the histogram.

Hereinafter, a specific example of noise level estimation in the noise level estimation unit 5 will be described.
The signal component of the left image SL is S1, and the signal component of the right image SR is Sr. The noise components of random noise on both images are equivalent to a normal distribution N (0, σ ² ) having an average value of 0 and a variance value σ ² . In this case, the luminance value Yl of the left image and the luminance value Yr of the right image can be expressed as follows.
Yl = Sl + N (0, σ ² )
Yr = Sr + N (0, σ ² )

Assuming that only the part where S1≈Sr can be measured by excluding the measurement, the dispersion value measured by the histogram is
Since Y1−Yr = S1−Sr + N (0,2σ ² ) ≈N (0,2σ ² ),
A histogram for the value of 2σ ² is obtained. For this reason, as the noise level estimation, a method can be considered in which the variance value having the highest frequency in the histogram is used as the most likely variance value to obtain the noise level of the entire screen. Another possible method is to use the center of gravity of the histogram as the noise level of the entire screen.

  The noise level estimation unit 5 calculates the noise level reliability simultaneously with the noise level estimation. If random noise can be accurately measured, the histogram will have a normal distribution shape. For this reason, reliability can be calculated | required from normal distribution likeness. As a specific method, there is a method using kurtosis or skewness. For example, the reliability is lowered when the frequency of the histogram is extremely concentrated on one portion or when the shape of the histogram distribution is distorted. Furthermore, the reliability of the noise level can be calculated based on the data of the measurement suitability determination result from the measurement suitability determination unit 3. For example, when the number of small blocks 31 determined not suitable for noise level measurement is extremely large, a method of reducing the reliability of the noise level is conceivable. Finally, the obtained noise level reliability can be multiplied by the noise level itself, and if the reliability is low, correction can be made to lower the noise level. The reliability of the noise level can also be used for an image processing apparatus that uses the output from the noise level measuring apparatus 1. For example, in an image processing apparatus including a circuit block controlled depending on a noise level, it can be used for reducing the noise level dependency. The processing by the noise level estimation unit 5 as described above can also be performed by a CPU (Central Processing Unit).

  According to the noise level measurement device 1 according to the present embodiment, a difference variance value is obtained by adding only a value for a small block determined to be suitable for noise level measurement to a difference variance value between two parallax images. Since the noise level of random noise is estimated based on the histogram, the random noise included in the stereo image can be measured with high accuracy. In particular, since the measurement area of the noise level is limited according to the feature of the stereo image, it is possible to prevent the difference value of the signal component itself that is not noise from being erroneously recognized as the noise component.

<Second Embodiment>
Next, an image processing apparatus according to the second embodiment of the present invention will be described. In addition, the same code | symbol is attached | subjected to the substantially same component as the noise level measuring apparatus which concerns on the said 1st Embodiment, and description is abbreviate | omitted suitably.

[Configuration of image processing apparatus]
FIG. 9 shows a configuration example of the image processing apparatus according to the present embodiment. This image processing apparatus includes a stereo matching circuit 20, a noise level measurement unit 21, and a noise removal unit 22. The noise level measuring unit 21 is configured by the noise level measuring apparatus 1 shown in FIG.

  Based on the noise level measured by the noise level measurement unit 21, the noise removal unit 22 generates random noise for each of two parallax images (two left and right stereo images (left image SL, right image SR)). To be removed. The noise removing unit 22 includes a frame memory 17, a first time direction recursive filter 18, and a second time direction recursive filter 19. The first time direction recursive filter 18 removes random noise from the left image SL. The second time direction recursive filter 19 removes random noise from the right image SR. The first time direction recursive filter 18 and the second time direction recursive filter 19 are commonly input with noise level data and data indicating noise level reliability from the noise level measurement unit 21. It has become. The first time direction recursive filter 18 and the second time direction recursive filter 19 can perform on / off control as to whether or not to remove random noise according to the reliability of the noise level. ing. For example, when the reliability of the noise level is extremely low, it is possible to control to turn off the filter operation.

  The stereo matching circuit 20 receives the left image SL and the right image SR after the random noise is removed by the noise removing unit 22. The stereo matching circuit 20 recognizes a 3D object from the left image SL and the right image SR by a stereo matching method, and outputs the 3D depth information. In the stereo matching method, for example, one of the left and right images is fixed, the other image is moved left and right, a matching point is searched for in small blocks, and depth information is calculated from the amount of deviation between the left and right images.

  Instead of the stereo matching circuit 20, a circuit that generates a display image for 3D display may be provided. Then, a display image for 3D display may be generated and output based on the left image SL and the right image SR after the random noise is removed.

[Operation and effect of image processing apparatus]
In this image processing apparatus, based on the noise level calculated from the left image SL and the right image SR, the feedback rates of the time direction recursive filters 18 and 19 for the left image SL and the right image SR are controlled. Specifically, when the noise level is high, the feedback rates of the time direction recursive filters 18 and 19 are increased. Conversely, when the noise level is small, the feedback rate is lowered. In this way, the feedback rate, that is, the intensity of noise removal is controlled according to the noise level.

  In this image processing apparatus, independent filters 18 and 19 are used for the left image SL and the right image SR, and at the same intensity on the left and right using a common noise level calculated from the left and right images SL and SR. Perform noise removal processing. This makes it possible to minimize the difference between the two images during noise removal. When generating a display image for 3D display or acquiring 3D depth information, it is possible to prevent an unnatural result due to an unbalance between the left and right images SL and SR.

  In recent years, two cameras have been used for crime prevention and in-vehicle applications, and two parallax images (two left and right stereo images) taken at the same time are obtained from each camera, and stereoscopic image recognition is performed by the stereo method. An increasing number of devices are available. This image processing apparatus can be used for such applications. It is considered that security cameras and in-vehicle cameras used for such purposes are often photographed in situations where the photographing conditions are poor and noise components are likely to be mixed, such as photographing in a dark place. According to this image processing apparatus, even in a video shot under such an adverse condition, a robust object recognition performance is provided by correctly removing a noise component according to the calculated noise level. It becomes possible. On the other hand, when the noise is low, it is possible to ensure more accurate object recognition performance by turning off the time direction recursive filters 18 and 19.

  By the way, there are cases where the characteristics and conditions of the camera at the time of shooting differ greatly between the left and right cameras, and there are extremely few background areas common between the left and right images. In such a case, it is considered that most of the region is excluded by the measurement exclusion function in the noise level measurement unit 21 and the noise level cannot be measured. In such a case, the noise level reliability output from the noise level measuring unit 21 may be used. For example, when the reliability is low, the time direction recursive filters 18 and 19 can be turned off to bring the entire system in a more stable direction.

  As described above, according to the image processing apparatus according to the present embodiment, random noise can be favorably removed from a stereo image. Thereby, for example, when performing stereoscopic image recognition by the stereo method, it is possible to suppress deterioration in object recognition performance even for a stereo image obtained under a bad shooting condition such as a security camera or an in-vehicle camera.

  SL ... Left image, SR ... Right image, 1 ... Noise level measuring device, 2 ... Differential variance calculation unit, 3 ... Measurement appropriateness determination unit, 4 ... Histogram generation unit, 5 ... Noise level estimation unit, 6 ... Differentiator, 7 ... Intra-block variance calculating unit, 8 ... First in-block average calculating unit, 9 ... Second in-block average calculating unit, 10 ... First edge detecting unit, 11 ... Second edge detecting unit, 12 ... Stereo matching circuit, 13 ... determining unit, 14 ... peak determining unit, 15 ... reliability calculating unit, 16 ... multiplier, 17 ... frame memory, 18 ... first time direction recursive filter, 19 ... second time direction Cyclic filter, 20 ... stereo matching circuit, 21 ... noise level measurement unit, 22 ... noise removal unit, 31 ... small block, 31L ... small left image block, 31R ... small right image block, 32L, 32R ... object, 2 ... object portion, 33 ... background image portion.

Claims (9)

Two parallax images having parallax with each other are input, and each parallax image is divided into a plurality of small blocks, and each of the plurality of small blocks is determined to be suitable for noise level measurement. And
A difference variance value calculation unit for obtaining a difference variance value between the two parallax images in units of the small blocks;
Of the difference variance values obtained by the difference variance value calculation unit, only a value for a small block determined to be suitable for noise level measurement by the measurement appropriateness determination unit is added to generate a histogram of the difference variance value A histogram generator to
A noise level measurement apparatus comprising: a noise level estimation unit that analyzes the histogram generated by the histogram generation unit and estimates a noise level of random noise included in common in the two parallax images. The noise level estimation unit calculates the reliability of the estimated noise level based on the normal distribution likelihood of the histogram and the information of the small block determined not suitable for noise level measurement by the measurement appropriateness determination unit. The noise level measurement apparatus according to claim 1, wherein the noise level measurement apparatus is configured to perform the noise level measurement. The noise level measurement device according to claim 2, wherein the noise level estimation unit corrects the estimated noise level according to the reliability of the noise level. The noise level estimation unit is configured to estimate the noise level based on a variance value having the highest frequency in the histogram or a variance value serving as a center of gravity in the histogram. The noise level measuring device according to item. The measurement appropriateness determination unit includes an edge detection unit that performs edge detection for each of the two parallax images in units of the small blocks, and determines that a small block including an edge component is not suitable for noise level measurement. The noise level measuring apparatus according to any one of claims 1 to 3, wherein the noise level measuring apparatus is configured as described above. The measurement appropriateness determining unit includes an intra-block average calculating unit that calculates an average of luminance in blocks in units of the small blocks for each of the two parallax images, and is in a position corresponding to the two parallax images. The noise level measurement according to any one of claims 1 to 3, wherein when the average luminance value of the small block is greatly different, it is determined that the small block is not suitable for noise level measurement. apparatus. The measurement appropriateness determination unit includes a stereo matching circuit that performs image analysis by a stereo matching method on the two parallax images, and determines background likelihood in units of the small blocks based on the image analysis by the stereo matching method. The noise level measuring apparatus according to any one of claims 1 to 3, wherein a small block corresponding to a background-like portion is determined to be suitable for noise level measurement. A noise level measurement unit that measures a noise level of random noise included in common to two parallax images having parallax with each other;
A noise removing unit that removes the random noise for each of the two parallax images based on the measured noise level;
The noise level measurement unit
The two parallax images are input, the parallax image is divided into a plurality of small blocks, and a measurement appropriateness determination unit that determines whether each of the plurality of small blocks is suitable for noise level measurement;
A difference variance value calculation unit for obtaining a difference variance value between the two parallax images in units of the small blocks;
Of the difference variance values obtained by the difference variance value calculation unit, only a value for a small block determined to be suitable for noise level measurement by the measurement appropriateness determination unit is added to generate a histogram of the difference variance value A histogram generator to
An image processing apparatus, comprising: a noise level estimation unit that analyzes the histogram generated by the histogram generation unit and estimates a noise level of random noise included in common in the two parallax images. The noise level estimation unit calculates the reliability of the estimated noise level based on the normal distribution likelihood of the histogram and the information of the small block determined not suitable for noise level measurement by the measurement appropriateness determination unit. Was made to
The image processing apparatus according to claim 8, wherein the noise removing unit performs on / off control on whether or not to remove the random noise according to reliability of the noise level.

Images