TWI506590B - Method for image noise reduction - Google Patents

Description

Method of removing image noise

The present invention relates to an image processing technique, and more particularly to a method of removing image noise.

In image processing, it is the most basic work to clearly present edges and enhance edges while removing unnecessary noise. In the process of shooting, digital images often generate noise due to the electronic components of the camera itself or external factors such as temperature and light. In order to cope with these noise problems, many methods for filtering noise have been proposed. However, the biggest problem with noise filtering is that it reduces the sharpness of the image. In other words, noise filtering can also result in a blurred image. In general, such situations apply edge-enhanced algorithms to improve image sharpness. However, edge-enhanced image processing often enhances the noise signal at the same time.

In the existing methods for removing image noise, some methods are used to use a median filter, a mean filter, or a low pass filter (LPF) to image noise. Remove. The above methods use an average operation of pixels in the entire image to remove image noise, without considering smooth areas and detail areas in the image. Another existing practice is Using a bilateral filter, the intensity of noise removal can be adjusted by adjusting the weight value. However, if the weight value is too high, it may cause blurred image problems; if the weight value is too low, it may reduce the ability to remove noise, so how to choose between removing noise and retaining image details. It is an important issue. In addition, in high-noise images, the effect of simply using bilateral filters to improve image quality is limited.

The invention provides a method for removing image noise, which can effectively filter out noise in an image and effectively preserve image details and edge information while removing image noise.

The method for removing image noise of the present invention comprises the following steps. The image to be processed is received first, wherein the image to be processed includes a plurality of pixels to be processed. One of the pixels to be processed is selected as the target pixel, and texture analysis is performed on the target pixel to determine whether the target pixel is located in a shading area. When the target pixel is not located in the shaded area, a trilateral noise reduction process is performed on the target pixel to generate a processed pixel.

In an embodiment of the invention, the method for removing image noise further includes performing a bilateral noise reduction process on the target pixel to generate a processed pixel when the target pixel is located in the shadow region. .

In an embodiment of the invention, the three-sided down is performed on the target pixel. The step of low noise processing includes first selecting a first core centered on the target pixel and selecting a second core centered on each reference pixel. Then, each pixel in the first core and each corresponding pixel in the second core are calculated to obtain a similarity value of each reference pixel corresponding to the target pixel. Each reference pixel further determines a corresponding pixel weight value according to each similarity value. Finally, the pixel values of the respective reference pixels in the mask belonging to the target pixel are multiplied by the corresponding pixel weight values to obtain the processed pixels.

In an embodiment of the invention, the formula for calculating the similarity value of each pixel in the first core and each corresponding pixel in the second core to obtain the similarity value of each reference pixel corresponding to the target pixel is: l = s + t × M. Where, Sim _j is the similarity value of the reference pixel, and w _c is the confidence weight value. For the distance weight value, For the first core centered on the target pixel, It is the second core centered on the reference pixel.

In an embodiment of the invention, the confidence weight value w _c is determined by the distance between the reference pixel and the target pixel.

In an embodiment of the invention, the step of performing bilateral noise removal processing on the target pixel includes first selecting a mask centered on the target pixel. Next, a distance weight value and an intensity closeness weight value corresponding to each of the reference pixels in the mask are calculated. And, each reference pixel is operated according to each distance weight value and each proximity intensity weight value to obtain a processed pixel.

In an embodiment of the invention, the formula for calculating the respective reference pixels according to the respective distance weight values and the respective proximity intensity weight values is: , j = s + t × N . Where P _i is the target pixel and P _j is the reference pixel. The distance weight value of the reference pixel, Is the proximity intensity weight value of the reference pixel.

In an embodiment of the present invention, the block sizes of the first and second cores are M×M, and the block size of the mask is N×N, where M<N and M and N are greater than 0. Positive integer.

Based on the above, the present invention proposes an integrated process architecture to remove image noise to improve image quality. Among them, by performing texture analysis on the image, the adaptive selection uses the bilateral noise removal or the three-sided noise removal algorithm to preserve the image details while removing the image noise.

The above described features and advantages of the invention will be apparent from the following description.

300, 500‧‧‧ mask

P _i ‧‧‧target pixel

P _j ‧‧‧ reference pixels

K ¹ ‧ ‧ first core

K ² ‧‧‧Second core

S110~S150‧‧‧Steps for removing image noise

S210~S230‧‧‧Steps for performing bilateral noise removal processing methods

S410~S440‧‧‧Steps to perform three-sided noise reduction processing

FIG. 1 is a flow chart of a method for removing image noise according to an embodiment of the invention.

2 is a flow chart of a method for performing bilateral noise removal processing on a target pixel according to an embodiment of the invention.

FIG. 3 is a schematic diagram of a mask according to an embodiment of the invention.

FIG. 4 is a flow chart of a method for performing three-sided noise reduction on a target pixel according to an embodiment of the invention.

FIG. 5 is a simplified schematic diagram of a mask and a core thereof according to another embodiment of the invention.

The present invention proposes an integrated process architecture to remove image noise problems. It adaptively uses a bilateral noise removal algorithm based on image texture information or uses a modified non-local mean noise removal algorithm, while retaining the advantages of both algorithms, thereby removing images At the same time as the noise, it achieves the effect of retaining the details of the image. In order to clarify the content of the present invention, the following examples are given as examples in which the present invention can be implemented. The examples are presented for illustrative purposes only and are not intended to limit the scope of the invention.

FIG. 1 is a flow chart of a method for removing image noise according to an embodiment of the invention. Referring to FIG. 1, the method flow of this embodiment is suitable for use in an image acquisition device such as a digital camera, a Digital Single Lens Reflex (DSLR) camera, a Digital Video Camcorder (DVC), or other image processing. Electronic devices such as smart phones and tablets are not limited to the above.

First, in step S110, the image to be processed is first received, wherein the image to be processed includes a plurality of pixels to be processed. Next, in step S120, one of the pixels to be processed is selected as the target pixel, and the target pixel is subjected to texture segmentation. Texture analysis processing. The texture analysis processing described herein may be applied by an ordinary knowledge in the art to select an existing texture analysis algorithm, and thus is not limited herein.

Next, in step S130, based on the result of the texture analysis, it is determined whether the target pixel is located in a shading area. The shaded regions described herein represent that the image has fewer contours or edges in this region and, therefore, may also be referred to as smooth regions. In other words, if the target pixel is in the shaded area, it can also be called a shading pixel.

In step S140, when the target pixel is located in the shaded area, a bilateral noise removal process may be performed on the target pixel by using a bilateral filter to generate the processed pixel. Conversely, in step S150, when the target pixel is not located in the shaded area, the improved non-local means filter may be used to perform trilateral noise reduction on the target pixel. Processed to produce processed pixels.

Accordingly, the method provided by the embodiment adaptively uses two different filters based on the result of the texture analysis, and can remove the image noise while achieving the effect of retaining the image detail information.

The following is a detailed description of the (improved) three-sided noise removal algorithm performed by the bilateral noise removal algorithm performed by the bilateral filter and the improved non-local averaging filter.

It should be noted that the filters for removing image noise can be roughly classified into two types, one is a local average filter and the other is a non-local average filter. among them, The bilateral noise removal algorithm performed by the bilateral filter is a common and effective local average filter. However, whether it is a non-local averaging filter or a bidirectional filter, the characteristics of the Gaussian filter are used to eliminate the noise problem in the image.

2 is a flow chart of a method for performing bilateral noise removal processing on a target pixel according to an embodiment of the invention. 2 is a detailed embodiment of step S140 of FIG. 1.

Referring to FIG. 2, in step S210, a mask centered on the target pixel is first selected. In this embodiment, the block size of the mask is N×N, and N is a positive integer greater than 0. That is to say, in this step, a two-dimensional square pixel array is selected with the target pixel as a center circle to perform the operation of removing noise. For example, FIG. 3 is a schematic diagram of a mask according to an embodiment of the invention. Referring to FIG. 3, the mask 300 is, for example, a 5×5 (N=5) array including one target pixel P _i and 24 reference pixels P _j .

Next, in step S220, a distance weight value and an intensity closeness weight value corresponding to each of the reference pixels in the mask are calculated. In detail, the formula for calculating the respective reference pixels according to the respective distance weight values and the respective proximity intensity weight values is as shown in the following formula (1): Where P _i is the target pixel and P _j is the reference pixel. The distance weight value of the reference pixel, Is the proximity intensity weight value of the reference pixel. The value of the index j is the calculated value of s + t × N.

The bidirectional filter reconstructs each pixel in the image to be processed using two weight values related to distance and proximity strength (similarity). Thus, the distance weight value refers to a higher reference value from the target pixel P _i closer to the reference pixel P _j, such that upon the basis of each reference pixel reconstruction target pixel P _i, closer to the reference target pixel P _i to P _j Distance weight value The higher. Right and approach the strength of the similarity values associated weight refers to the periphery of the target pixel P _i P _j in respective reference pixels, the more similar to the target pixel P _i P _j of the reference pixels having the higher reference value, so that in accordance with various reference when the target pixel P _j reconstructed pixel P _i, P _j P _i and the target pixel like the reference pixel is close to a weight value of the intensity weights The higher.

After the distance weight value and the proximity strength weight value of each reference pixel are obtained, step S230 may be continued, and each reference pixel is operated according to each distance weight value and each proximity intensity weight value to obtain a processed pixel (ie, processed by operation) After the target pixel).

4 is a flow chart of performing (improved) three-sided noise reduction processing on a target pixel according to an embodiment of the invention. 4 is a detailed embodiment of step S150 of FIG. 1.

Referring to FIG. 4, in step S410, a first core centered on the target pixel is selected, and a second core centered on each reference pixel is selected. Different from the bidirectional filter, the weight value is determined for each reference pixel. The three-side noise reduction method is used to check the similarity of a smaller size mask around the target pixel centered on the target pixel. degree. The smaller size mask is referred to as the "core" in this embodiment. In this embodiment, the block size of the mask is N×N, and the block sizes of the first and second cores are N×N, where M<N and M and N are Is a positive integer greater than zero.

Next, in step S420, each pixel in the first core and each corresponding pixel in the second core are calculated to obtain a similarity value of each reference pixel corresponding to the target pixel. For example, FIG. 5 is a simplified schematic diagram of a mask and its core according to another embodiment of the invention. Referring to FIG. 5, the mask 500 is centered on the target pixel P _i , and the block surrounded by the target pixel P _i is, for example, the first core K ¹ , and the block surrounded by the reference pixel P _j is, for example, the second core. K ² .

Wherein, the formula for calculating the similarity value of each pixel in the first core and each corresponding pixel in the second core to obtain the similarity value of each reference pixel corresponding to the target pixel is as shown in the following formula (2): Wherein, Sim _j is a similarity value of the reference pixel P _j , and w _c is a confidence weight value, For the distance weight value, For the first core centered on the target pixel P _i , It is a second core centered on the reference pixel P _j . The value of the index l is the calculated value of s + t × M.

The similarity value Sim _j is measured by similar measures by two cores. If the similarity value Sim _{j is} higher, the lower the similarity is. Conversely, if the similarity value Sim _{j is} lower, the higher the similarity is. It should be particularly noted that, in the formula (2) for calculating the similarity value, the present invention uses the confidence weight value w _c to adjust the intensity of removing noise. The confidence weight value w _c is determined by the distance between the reference pixel P _j and the target pixel P _i .

In step S430, each reference pixel further determines a corresponding pixel weight value according to each similarity value. Finally, in step S440, the mask belonging to the target pixel is The pixel values of the respective reference pixels within are multiplied by the corresponding pixel weight values to obtain processed pixels. Accordingly, the present invention improves the image quality by improving the method of calculating the similarity value so that the three-side noise reduction processing method can improve the image quality compared with the existing non-local average algorithm.

In summary, the method for removing image noise according to the present invention adaptively uses a bilateral noise removal algorithm or a three-side noise reduction method according to image texture information, while retaining the advantages of both algorithms. According to the purpose of removing image noise, it can achieve the effect of retaining the details of the image without causing the problem of blurred images. In addition, the present invention improves the image quality by improving the method of calculating the similarity value so that the three-side noise reduction processing method can improve the image quality compared with the existing non-local average algorithm. The image acquisition device using the method for removing image noise according to the present invention can effectively improve the image output quality at high sensitivity.

Although the present invention has been disclosed in the above embodiments, it is not intended to limit the present invention, and any one of ordinary skill in the art can make some changes and refinements without departing from the spirit and scope of the present invention. The scope of the invention is defined by the scope of the appended claims.

S110~S150‧‧‧Steps for removing image noise

Claims (8)

A method for removing image noise includes: receiving a to-be-processed image, wherein the image to be processed includes a plurality of pixels to be processed; selecting one of the pixels to be processed as a target pixel, and performing a target pixel A texture analysis is performed to determine whether the target pixel is located in a shaded area; and when the target pixel is not located in the shaded area, a Trilateral noise removal process is performed on the target pixel to generate a processed pixel. The method for removing image noise according to claim 1, further comprising: performing a bilateral noise removal process on the target pixel when the target pixel is located in the shadow region to generate a process. Rear pixel. The method for removing image noise according to claim 1, wherein the step of performing the three-side noise reduction processing on the target pixel comprises: respectively selecting a first core (Kernel) centered on the target pixel And selecting a second core centered on each of the reference pixels; calculating each pixel in the first core and each corresponding pixel in the second core to obtain each of the reference pixels corresponding to the target a similarity value of the pixel; each of the reference pixels determines a corresponding one-pixel weight value according to each similarity value; and multiplies the pixel value of each of the reference pixels in a mask of the target pixel by a pair The pixel weight value should be taken to obtain the processed pixel. The method for removing image noise according to claim 3, wherein each of the pixels in the first core and the corresponding pixels in the second core are calculated to obtain that each of the reference pixels corresponds to The similarity value of the target pixel is expressed by the following formula (1): Where, Sim _j is the similarity value of the reference pixel j, and w _c is a confidence weight value. For the distance weight value, For the first core centered on the target pixel i, The second core centered on the reference pixel j. The method for removing image noise according to claim 4, wherein the confidence weight value w _c is determined by a distance between the reference pixel j and the target pixel i. The method for removing image noise according to claim 2, wherein the step of performing the bilateral noise removal processing on the target pixel comprises: selecting a mask centered on the target pixel; calculating the mask Each of the reference pixels corresponds to a distance weight value of the target pixel and a proximity strength weight value; and each of the reference pixels operates according to each of the distance weight values and each of the proximity intensity weight values to obtain the processed pixel. The method for removing image noise according to claim 6, wherein each of the reference pixels is calculated according to each of the distance weight values and each of the proximity strength weight values is as follows: (2): Wherein P _{i is} the target pixel i, and P _{j is} the reference pixel j, The distance weight value of the reference pixel j, The proximity strength weight value of the reference pixel j. The method for removing image noise according to claim 3, wherein the block size of the first core and the second core is M×M, and the block size of the mask is N×N, wherein M< N and M and N are both positive integers greater than zero.