Image detail enhancement and its ASIC implementation method and system

Abstract

This invention relates to the field of image processing technology, specifically to an image detail enhancement method and system and its ASIC implementation, comprising the following steps: S1 Initialization, inputting the image to be processed, and performing parameter-configurable weighted mean filtering on the input image; S2 Calculating the difference between the current center pixel and the mean filter output, and obtaining the difference gain coefficient and grayscale gain coefficient from a table; S3 Combining the weighted mean filter output from S1, the difference output from S2, and the two gain coefficients, calculating the detail enhancement output result. This invention can adjust the detail gain according to the pixel value of the current pixel, effectively solving the overshoot and undershoot problems in image detail enhancement; it can select different filter coefficients and gain curves according to different application scenarios, and performs detail enhancement only in the Y channel in the YUV domain, effectively reducing the computational load and facilitating ASIC implementation.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, specifically to an image detail enhancement method and system and its ASIC implementation. Background Technology

[0002] During image acquisition and transmission, images are affected by the imaging environment, imaging equipment, and transmission system, resulting in noisy images and difficulty in discerning details, which greatly hinders the accuracy of subsequent image processing. Image detail enhancement aims to improve image quality by purposefully enhancing the recognizability of certain target information according to the requirements of subsequent image processing, expressing the features of the target of interest in the image in more detail, while suppressing noise.

[0003] Existing technologies cannot adjust detail gain based on the current pixel value; cannot solve the problems of overshoot and undershoot; cannot select different filter coefficients according to different application scenarios; and have excessive computational complexity, making them unsuitable for ASIC implementation. Therefore, this paper presents an image detail enhancement method and system with ASIC implementation. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention discloses an image detail enhancement method and system for ASIC implementation, which solves the problems that existing technologies cannot adjust detail gain according to the current pixel value; cannot solve the problems of "overshoot" and "undershoot"; cannot select different filtering coefficients according to different application scenarios; and have excessive computational load, which is not conducive to ASIC implementation.

[0005] This invention is achieved through the following technical solution:

[0006] In a first aspect, the present invention provides an image detail enhancement method and its ASIC implementation method, comprising the following steps:

[0007] S1 initialization: Input the image to be processed and perform parameter-configurable weighted mean filtering on the input image;

[0008] S2 calculates the difference between the current center pixel and the mean filter output, and looks up the difference gain coefficient and grayscale gain coefficient in the table;

[0009] S3 combines the weighted mean filter output from S1, the difference output from S2, and two gain coefficients to calculate the detailed enhanced output result.

[0010] Furthermore, in the method, during weighted mean filtering, the mean filtering coefficients are obtained differently depending on the position of the pixel within the filtering window, and the filtering coefficients are adjusted in real time according to different application scenarios.

[0011] Furthermore, in the method, during weighted mean filtering, all filtering coefficients are integer multiples of 2, and in the ASIC implementation, shifting is used instead of multiplication.

[0012] Furthermore, in the method described above, the calculation process for weighted mean filtering is as follows:

[0013]

[0014] M = W SUM ∑P J W J

[0015] W SUM =1 / ∑W J

[0016] Among them, X J W represents the shift value for each pixel within the filtering window. J These are the filtering coefficients for each pixel within the filtering window.

[0017] Furthermore, in the method, the calculation process for the filtered difference is as follows:

[0018] L DIFF =PM

[0019] Where P is the center pixel within the filtering window.

[0020] Furthermore, the method uses an unsharpened mask, first applying a high-pass filter and multiplying it by a scaling factor, then adding the result to the original image to enhance high-frequency information.

[0021] Furthermore, in the method, the difference gain coefficient FD and grayscale gain coefficient FI are obtained by looking up a table, and the output of detail enhancement is:

[0022] Y = Y + L DIFF *FI*FD.

[0023] Secondly, the present invention provides an image detail enhancement system and its ASIC implementation system. The system is used to implement the image detail enhancement and its ASIC implementation method described in the first aspect, and includes an image caching module, a data selector, a grayscale gain coefficient lookup table module, a configurable weighted average filtering module, a difference gain coefficient lookup table module, and an output calculation module.

[0024] Furthermore, the image caching module is used to cache image data and simultaneously generate a 7x7 neighborhood of the current pixel.

[0025] The data selector is used to select the center pixel and its neighboring pixel data that participate in detail enhancement;

[0026] The grayscale gain coefficient lookup table module is used to look up the grayscale gain coefficient based on the Y channel value of the current pixel.

[0027] The configurable weighted mean filtering module is used to calculate the filtered pixel values ​​and the difference between the original image and the filtered image.

[0028] The difference gain coefficient lookup table module is used to look up the difference gain coefficient based on the image difference obtained by the configurable weighted mean filter.

[0029] Output calculation is used to calculate the output image based on the Y channel value of the center pixel, the image difference, the grayscale gain coefficient, and the difference gain coefficient.

[0030] Furthermore, the entries in the grayscale gain coefficient lookup table module can be configured by the CPU or other main controller;

[0031] The filtering coefficients of the configurable weighted average filtering module can be configured by the CPU or other main controller;

[0032] The table entries of the difference gain coefficient lookup module can be configured by the CPU or other main controller.

[0033] The beneficial effects of this invention are as follows:

[0034] This invention can adjust the detail gain according to the pixel value of the current pixel, effectively solving the problems of "overshoot" and "undershoot" in image detail enhancement; this invention can select different filtering coefficients and different gain curves according to different application scenarios, and perform detail enhancement only in the Y channel in the YUV domain, effectively reducing the amount of computation and facilitating ASIC implementation. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a diagram illustrating the image detail enhancement method and its ASIC implementation method of the present invention;

[0037] Figure 2 This is a lookup table diagram of the difference gain coefficient in an embodiment of the present invention;

[0038] Figure 3 This is a table lookup diagram of grayscale gain coefficients in an embodiment of the present invention;

[0039] Figure 4 This is a system diagram of image detail enhancement and its ASIC implementation according to an embodiment of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Example 1

[0042] Reference Figure 1 As shown, this embodiment provides an image detail enhancement method and its ASIC implementation, including the following steps:

[0043] S1 initialization: Input the image to be processed and perform parameter-configurable weighted mean filtering on the input image;

[0044] S2 calculates the difference between the current center pixel and the mean filter output, and looks up the difference gain coefficient and grayscale gain coefficient in the table;

[0045] S3 combines the weighted mean filter output from S1, the difference output from S2, and two gain coefficients to calculate the detailed enhanced output result.

[0046] This embodiment can adjust the detail gain according to the pixel value of the current pixel; it can effectively solve the problems of "overshoot" and "undershoot" in image detail enhancement.

[0047] This embodiment allows for the selection of different filter coefficients and gain curves based on different application scenarios.

[0048] This embodiment performs detail enhancement only in the Y channel within the YUV domain, effectively reducing computational load and facilitating ASIC implementation.

[0049] Example 2

[0050] At the implementation level, this embodiment provides a weighted mean filtering and filter difference calculation. In this embodiment, image detail enhancement is usually performed on the three RGB components in the RGB domain. Although this detail enhancement method can achieve good results, it has a large amount of computation and is prone to color distortion.

[0051] Therefore, this embodiment performs detail enhancement in the YUV domain, and only in the Y channel, which can effectively reduce the amount of computation and avoid color distortion.

[0052] W0 W1 W2 W3 W2 W1 W0 W1 W4 W5 W6 W5 W4 W1 W2 W5 W7 W8 W7 W5 W2 W3 W6 W8 W9 W8 W6 W3 W2 W5 W7 W8 W7 W5 W2 W1 W4 W5 W6 W5 W4 W1 W0 W1 W2 W3 W2 W1 W0

[0053] In this scheme, as shown in the table above, the window size for the weighted mean filter is 7x7, and the coefficients of each pixel within the window are shown in the table above. The calculation process for the weighted mean filter is as follows:

[0054]

[0055] M = W SUM ∑P J W J

[0056] W SUM =1 / ∑W J

[0057] The calculation process for the filtered difference is as follows:

[0058] L DIFF =PM

[0059] Among them, X J W represents the shift value for each pixel within the filtering window. J Here, P represents the filtering coefficients for each pixel within the filtering window, and P is the center pixel within the filtering window.

[0060] In this embodiment, traditional mean filtering uses the same filtering coefficients for all pixels within the filtering window, and the filtering coefficients are fixed and will not be optimized to obtain different filtering coefficients according to different application scenarios.

[0061] Therefore, the mean filtering coefficients in this embodiment vary depending on the pixel's position within the filtering window, and can be adjusted in real-time according to different application scenarios, thus improving the system's robustness.

[0062] Since the filter coefficients in this embodiment are all integer multiples of 2, they can be replaced by shifting in the ASIC implementation, which further reduces the amount of computation.

[0063] Example 3

[0064] At the implementation level, this embodiment provides a gain coefficient lookup table and detail enhancement output. A classic image enhancement algorithm in this embodiment is the unsharp masking technique.

[0065] In this embodiment, high-pass filtering is first applied and multiplied by a scaling factor. The result is then added to the original image to enhance the high-frequency information.

[0066] Although this embodiment of USM is simple to implement and applicable to most application scenarios, the output image will have obvious "overshoot" or "undershoot" artifacts. In order to eliminate "overshoot" or "undershoot" artifacts, they must be properly controlled.

[0067] In this embodiment, the filter gain coefficient and grayscale gain coefficient are obtained by looking up tables based on the calculated filter difference and the pixel value of the current pixel. By controlling the intensity of detail enhancement through these two gain coefficients, the problems of "overshoot" or "undershoot" can be effectively solved.

[0068] The table diagram for the difference gain coefficient in this embodiment is shown below. Figure 2 As shown in the table, the grayscale gain coefficient lookup table is as follows: Figure 3 As shown, based on the two curves mentioned above, the difference gain coefficient FD and the grayscale gain coefficient FI can be obtained by looking up the table.

[0069] from Figure 2 and 3 As can be seen, when the difference and grayscale value are too large or too small, both gain coefficients are constant and small values, thus effectively solving the problems of "overshoot" or "undershoot". At the same time, in the [0,1] region, as the difference and grayscale value increase, the gain coefficient also increases, which can effectively enhance image details. In the [1,3] interval, as the difference and grayscale value increase, the gain coefficient gradually decreases to prevent "overshoot".

[0070] Ultimately, the output of the enhanced detail is:

[0071] Y = Y + L DIFF *FI*FD.

[0072] Example 4

[0073] Reference Figure 4 As shown, this embodiment provides an image detail enhancement system and its ASIC implementation system, including an image caching module, a data selector, a grayscale gain coefficient lookup table module, a configurable weighted average filtering module, a difference gain coefficient lookup table module, and an output calculation module.

[0074] The image caching module in this embodiment is used to cache image data and generate a 7x7 neighborhood of the current pixel.

[0075] The data selector in this embodiment is used to select the data of the center pixel and its neighboring pixels that participate in detail enhancement.

[0076] The grayscale gain coefficient lookup table module in this embodiment is used to look up the grayscale gain coefficient based on the Y channel value of the current pixel.

[0077] This embodiment can be configured with a weighted average filtering module to calculate the filtered pixel values ​​and the difference between the original image and the filtered image;

[0078] The difference gain coefficient lookup table module in this embodiment is used to look up the difference gain coefficient in a table based on the image difference obtained by the configurable weighted mean filter.

[0079] Output calculation is used to calculate the output image based on the Y channel value of the center pixel, the image difference, the grayscale gain coefficient, and the difference gain coefficient.

[0080] In this embodiment, the table entries of the grayscale gain coefficient lookup module can be configured by the CPU or other main controller;

[0081] The filter coefficients of the weighted average filter module configured in this embodiment can be obtained by the CPU or other main controller;

[0082] In this embodiment, the table entries of the difference gain coefficient lookup module can be configured by the CPU or other main controller.

[0083] In summary, this invention can adjust the detail gain according to the pixel value of the current pixel, effectively solving the problems of "overshoot" and "undershoot" in image detail enhancement; this invention can select different filtering coefficients and different gain curves according to different application scenarios, and perform detail enhancement only in the Y channel in the YUV domain, effectively reducing the amount of computation and facilitating ASIC implementation.

[0084] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An ASIC-based method of image detail enhancement, characterized by, The method comprises the following steps: S1 initialization, inputting an image to be processed, and performing parameter-configurable weighted mean filtering on the input image; S2 calculating the difference between the current center pixel point and the mean filtering output, and looking up the difference gain coefficient and the gray gain coefficient; S3 combining the weighted mean filtering output in S1, the difference output in S2, and the two gain coefficients to calculate the detail enhancement output result.

2. The ASIC-based image detail enhancement method of claim 1, wherein, In the method, when performing weighted mean filtering, the mean filtering coefficients are different according to the positions of the pixel points in the filtering window, and the filtering coefficients are adjusted in real time according to different application scenarios.

3. The ASIC-based image detail enhancement method of claim 2, wherein, In the method, when performing weighted mean filtering, the filtering coefficients are all integer multiples of 2, and the multiplication calculation is replaced by the shift method in ASIC implementation.

4. The ASIC-based image detail enhancement method of claim 2, wherein, In the method, the calculation process of the weighted mean filtering is as follows: M = W SUM ∑P J W J W SUM = 1 / ∑W J wherein X J is the shift value of each pixel point in the filter window, W J is the filter coefficient of each pixel in the filter window.

5. The ASIC-based image detail enhancement method of claim 4, wherein, In the method, the calculation process of the filtering difference is as follows: L DIFF = P - M Wherein, P is the center pixel point in the filtering window.

6. The ASIC-based image detail enhancement method of claim 1, wherein, In the method, the image to be processed uses a non-sharpening mask, first passes through a high-pass filter and is multiplied by a scaling coefficient, and the result is added to the original image to enhance the high-frequency information.

7. The ASIC-based image detail enhancement method of claim 1, wherein, In the method, the difference gain coefficient FD and the gray gain coefficient FI are obtained by looking up the table, and the output of the detail enhancement is: Y' = Y + L DIFF *FI*FD Y' = Y + L * (Y - Y') / Y DIFF is the filtered difference.

8. An ASIC-based image detail enhancement system for implementing the ASIC-based image detail enhancement method according to any one of claims 1 to 7, characterized by The system comprises an image buffer module, a data selector, a gray gain coefficient lookup table module, a configurable weighted mean filtering module, a difference gain coefficient lookup table module, and an output calculation module.

9. The ASIC-based image detail enhancement system according to claim 8, characterized in that, The image buffer module is used to buffer image data and generate a 7x7 neighborhood of the current pixel point; The data selector is used to select the center pixel point and its neighborhood pixel point data participating in the detail enhancement; The gray gain coefficient lookup table module is used to look up the gray gain coefficient according to the Y channel value of the current pixel point; The configurable weighted mean filtering module is used to calculate the filtered pixel value and the difference between the original image and the filtered image; The difference gain coefficient lookup table module is used to look up the difference gain coefficient according to the image difference obtained by the configurable weighted mean filter; The output calculation is used to calculate the output image according to the center pixel Y channel value, the image difference, the gray gain coefficient, and the difference gain coefficient.

10. The ASIC-based image detail enhancement system of claim 8, wherein, The table entries of the gray gain coefficient lookup table module can be configured by a CPU or other main controller; The filtering coefficients of the configurable weighted mean filtering module can be configured by a CPU or other main controller; The table entries of the difference gain coefficient lookup table module can be configured by a CPU or other main controller.

Images