JP2013005219A - Noise removal device, noise removal method, and noise removal program image enhancement device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a noise removal device which removes noise from an image, without regard to a filter size.SOLUTION: An original image is composed of plural pixels and each pixel has an original luminance value. The original image is transmitted to a calculator 101. The calculator 101 having received the original image applies gaussian filter to the original image to smooth the original luminance value. The smoothed original luminance value is called a transformed luminance value. A comparator 102 compares the original luminance value with the transformed luminance value and outputs a lower luminance value, that is a smaller luminance value, as an output luminance value. The output luminance value is transmitted to an image composer 103. The image composer 103 creates an output image from the output luminance value.

Description

  The present invention relates to a noise removal apparatus, a noise removal method, and a noise removal program that remove noise contained in an image.

  A configuration is known in which a median filter is applied to a color difference signal constituting an image, the color difference signal is corrected using a signal obtained thereby, and noise included in the image is removed (Patent Document 1). The median filter is a filter that selects a median value from color difference signals of a plurality of peripheral pixels existing around the pixel of interest and uses the median value as a color difference signal of the pixel of interest. The range of surrounding pixels is called the filter size.

JP 2010-136059 A

  However, depending on the size of the filter, noise may not be removed accurately. That is, if the filter size is reduced, noise may remain, and if the filter size is increased, the image may blur.

  The present invention has been made in view of these problems, and an object thereof is to obtain a noise removing device that removes noise of an image without depending on a filter size.

  The noise removal apparatus according to the first invention of the present application compares the original luminance value and the smoothed luminance value with a calculation unit that processes a plurality of original luminance values constituting an image to create a converted luminance value, And a comparator that outputs the luminance value as an output luminance value.

  The calculation unit preferably creates a converted luminance value by smoothing the original luminance value by applying a Gaussian filter to the original luminance value.

  The calculation unit may create a converted luminance value by smoothing the original luminance value by applying a bilateral filter to the original luminance value.

  In the noise removal method according to the second invention of the present application, the step of smoothing a plurality of original luminance values constituting an image to create a smoothed luminance value, and comparing the original luminance value and the smoothed luminance value, the lower luminance A step of outputting the value as an output luminance value; and a step of creating an output image composed of the output luminance value.

  The noise removal program according to the third invention of the present application smoothes a plurality of original luminance values constituting an image and causes the smoothed luminance value calculation unit to create a smoothed luminance value, and sets the original luminance value and the smoothed luminance value as luminance values. Comparing to the comparison unit, the method includes a step of causing the luminance value comparison unit to output the lower luminance value as an output luminance value and a step of causing the image composition unit to generate an output image composed of the output luminance values.

  According to the present invention, it is possible to obtain a noise removing device that removes image noise without depending on the filter size.

It is the block diagram which showed the noise removal apparatus schematically. It is the figure which showed the original luminance value which comprises an original image. It is the figure which showed the smoothing luminance value. It is the figure which showed the output luminance value. It is the flowchart which showed the synthetic | combination process. It is the flowchart which showed the smoothing process.

  Hereinafter, a noise removing apparatus 100 according to the present invention will be described with reference to the accompanying drawings. First, the configuration of the noise removal apparatus 100 will be described with reference to FIG.

  The noise removal apparatus 100 is provided as a part of the image processing apparatus 220, and includes a calculation unit 101, a comparison unit 102, and an image composition unit 103. The image processing apparatus 220 includes a video signal creation unit 221, a memory 222, and a video output adjustment unit 223, and is connected to the imaging unit 210 and the display 230. The imaging unit 210 is an endoscope scope or a camera, for example, and captures a subject image and outputs an image signal. The image processing device 220 receives an image signal from the imaging unit 210, performs image processing, and then outputs the image signal to the display 230. The display 230 receives an image signal from the image processing device 220 and displays an image.

  Next, the configurations of the noise removal device 100 and the image processing device 220 will be described. The image signal received from the imaging unit 210 is input to the video signal creation unit 221. The video signal creation unit 221 performs a predetermined process on the image signal to create an original image. The original image is composed of a plurality of pixels, and each pixel has an original luminance value. The created original image is transmitted to the memory 222 and the calculation unit 101. The memory 222 stores the original image.

  Receiving the original image, the calculation unit 101 applies a Gaussian filter to the original luminance value to smooth the original luminance value. The smoothed original luminance value is called a converted luminance value. The Gaussian filter is, for example, an n-by-n matrix, and the degree of smoothing can be adjusted by changing the value of σ and the value of n constituting the Gaussian function. The filter size can be changed by changing the value of n. The converted luminance value is transmitted to the comparison unit 102.

  The comparison unit 102 compares the original luminance value and the converted luminance value, and outputs the lower luminance value, that is, the smaller luminance value as the output luminance value. The output luminance value is transmitted to the image composition unit 103. The image composition unit 103 creates an output image from the output luminance value, and transmits the output image to the memory 222 and the video output adjustment unit 223. The memory 222 stores the output image, and the video output adjustment unit 223 converts the output image into a signal format that matches the format of the display 230 and outputs the converted image to the display 230.

  Next, a process in which the original luminance value is converted into the output luminance value through the converted luminance value will be described with reference to FIGS. 2 shows a graph of original luminance values, FIG. 3 shows a graph of converted luminance values, and FIG. 4 shows a graph of output luminance values.

  Referring to FIG. 2, there are three so-called colored noise portions. Colored noise refers to noise that appears in a color image and has a luminance value that is completely different from the luminance value of surrounding pixels. For example, when a yellow pixel surrounded by pink peripheral pixels exists, the yellow pixel is referred to as colored noise. When the original luminance value shown in FIG. 2 is smoothed by applying a Gaussian filter, the converted luminance value shown in FIG. 3 is obtained. When the original luminance value and the converted luminance value are compared, there is no protruding luminance value, in other words, there is no portion that changes rapidly.

  Next, the original luminance value and the converted luminance value are compared, and the smaller luminance value is adopted as the output luminance value. Thereby, the output luminance value shown in FIG. 4 is obtained. The converted luminance value obtained by smoothing tends to be smaller than the original luminance value in the colored noise portion and larger than the original luminance value in the other portions. Therefore, by comparing the original luminance value and the converted luminance value and adopting the smaller luminance value, colored noise can be reduced. That is, when the original luminance value and the output luminance value are compared in FIG. 4, the protruding luminance value, that is, the colored noise portion is eliminated, and the substantially luminance value is maintained in the other portions.

  Next, the synthesis process will be described with reference to FIG. The synthesis process is a process executed by the noise removal apparatus 100 and is executed when an original image is received from the video signal creation unit 221. Here, the original luminance value is Img1 (x, y), the original luminance value copy is Img2 (x, y), the converted luminance value is Img2blur (x, y), and the output luminance value is Img (x, y). To do.

  In the first step S51, the original luminance value Img1 (x, y) included in the original image is substituted for Img2 (x, y). In the next step S52, smoothing processing is performed on Img2 (x, y) to obtain a converted luminance value Img2blur (x, y). The smoothing process is a process of applying a Gaussian filter to the original luminance value, and details will be described later.

  In the next step S53, the original luminance value Img1 (x, y) is compared with the converted luminance value Img2blur (x, y). If the converted luminance value Img2blur (x, y) is smaller than the original luminance value Img1 (x, y), the process proceeds to step S54; otherwise, the process proceeds to step S55.

  In step S54, the converted luminance value Img2blur (x, y) is substituted into the output luminance value Img (x, y). In step S55, the original luminance value Img1 (x, y) is substituted into the output luminance value Img (x, y). By these processes, the smaller luminance value of the original luminance value Img1 (x, y) and the converted luminance value Img2blur (x, y) becomes the output luminance value Img (x, y).

  Next, the smoothing process will be described with reference to FIG. The smoothing process is a process for applying a Gaussian filter to the original luminance value. More specifically, this is a process of multiplying the luminance values of all the pixels existing in the same size range as the filter size by a matrix forming a Gaussian filter. Here, the filter size of the Gaussian filter is assumed to be Fsize. The filter size is determined in advance according to the subject of the image, the experience value, and the like, and is input to the noise removal apparatus 100.

  In the first step S61, a Gaussian filter Gauss (x, y) is created according to the input filter size. In the next step S62, -Fsize / 2 is substituted for the parameters i and j.

  In the next step S63, it is determined whether or not the parameter j is smaller than Fsize / 2. If so, the process proceeds to step S64; otherwise, the process ends.

  In step S64, -Fsize / 2 is substituted for the parameter i. In the next step S65, it is determined whether or not the parameter j is smaller than Fsize / 2. If so, the process proceeds to step S66; otherwise, the process proceeds to step S68 and increments the parameter j by one.

  In step S66, the parameter i is incremented by 1, and the process proceeds to step S67. In step S67, the product of the original luminance value copy Img2 (x, y) and the Gaussian filter Gauss (x, y) is obtained, and the result is substituted into the converted luminance value Img2blur (x, y). Then, the process proceeds to step S65.

  By repeating steps S63 to S68 as many times as the number of pixels included in the filter size, the converted luminance value Img2blur (x, y) is obtained.

  In the present embodiment, the original luminance value and the converted luminance value are compared, and the smaller luminance value is adopted. Therefore, noise that remains when the filter size is small can be removed. That is, noise can be reduced without depending on the filter size.

  Note that a bilateral filter may be used instead of the Gaussian filter. Further, the imaging unit 210 is not limited to an endoscope scope or a camera.

  The noise removing device 100 may be used as an independent device instead of being part of the image processing device 220.

DESCRIPTION OF SYMBOLS 100 Noise removal apparatus 101 Calculation part 102 Comparison part 103 Image composition part 210 Image pick-up part 220 Image processing apparatus 221 Video signal creation part 222 Memory 223 Video output adjustment part 230 Display

Claims (5)

A calculation unit that processes a plurality of original luminance values constituting an image to create a converted luminance value;
A noise removing apparatus comprising: a comparison unit that compares the original luminance value with the smoothed luminance value and outputs a lower luminance value as an output luminance value.   The noise removing apparatus according to claim 1, wherein the calculation unit smoothes the original luminance value by applying a Gaussian filter to the original luminance value to create a converted luminance value.   The noise removal apparatus according to claim 1, wherein the calculation unit smoothes the original luminance value by applying a bilateral filter to the original luminance value to create a converted luminance value. Smoothing a plurality of original luminance values constituting an image to create a smooth luminance value;
Comparing the original luminance value with the smoothed luminance value and outputting the lower luminance value as an output luminance value;
Creating an output image composed of the output luminance values. Smoothing a plurality of original luminance values constituting the image and causing the smoothed luminance value calculation unit to create a smoothed luminance value;
Comparing the original luminance value and the smoothed luminance value to a luminance value comparison unit, and outputting the lower luminance value to the luminance value comparison unit as an output luminance value;
A noise removal program comprising: causing an image composition unit to create an output image composed of the output luminance values.

Images