CN103297659A - Image edge processing method and image processing device - Google Patents

Abstract

The invention relates to an image edge processing method and an image processing device. In one embodiment of the method for edge processing of an image, a first directional gradient along a first direction and a second directional gradient along a second direction are calculated for a cross pattern centered at a target pixel in an input image. According to the first direction gradient and the second direction gradient, whether the target pixel is compensated based on the pixel values of a first plurality of pixels along the second direction in the cross-shaped pattern or the target pixel is compensated based on the pixel values of a second plurality of pixels along the first direction or the pixel value of the target pixel is output is determined.

Description

Image edge processing method and image processing device

technical field

The invention relates to an image edge processing method and an image processing device.

Background technique

In image processing, because the input image may be based on an analog gain as the analog gain or the result of selecting a larger ISO value (such as 400, 800 or above) or the result of shooting when the light sensitivity is insufficient, the input image may appear noisy ( Noise) or current gain to enhance the noise, resulting in uneven edges in the image. The traditional method of noise reduction, for example, uses a low-pass filter to process, which will make the processing result blurred, so edge enhancement processing is usually added afterwards to compensate. In this way, although the noise is blurred, it still exists in the image, which affects the quality of the image.

Contents of the invention

The invention relates to an image edge processing method and an image processing device.

According to an aspect of the present invention, an embodiment of an image edge processing method is proposed. An embodiment of the method includes the following steps. (a) In an area of an input image, for a cross-shaped pattern centered on a target pixel in the area, calculate a first directional gradient along a first direction and a second directional gradient along a second direction . (b) judging whether the gradient in the first direction is greater than a first comparison gradient value, wherein the first comparison gradient value is determined according to the gradient in the second direction. (c) if the determination result of step (b) is yes, then based on the pixel values of the first plurality of pixels along the second direction in the cross-shaped pattern, a first estimated pixel value is obtained as a value of the target pixel Estimated pixel values. (d) If the judgment result of step (b) is negative, it is judged whether the gradient in the second direction is greater than a second comparison gradient value, wherein the second comparison gradient value is determined according to the gradient in the first direction. (e) If the judgment result of step (d) is yes, then based on the pixel values of the second plurality of pixels along the first direction in the cross-shaped pattern, a second estimated pixel value is obtained as the estimated pixel value of the target pixel measure the pixel value. (f) If the judgment result of step (d) is no, then output the pixel value of the target pixel.

According to another aspect of the present invention, an embodiment of an image processing device is proposed. The communication device includes a noise suppression unit, an edge low-pass filter unit and an edge enhancement unit. The noise suppression unit includes a low-pass filter, receives a first image, and performs a noise suppression process to output a second image. The edge low-pass filter unit receives the second image, performs edge low-pass filter processing to smooth the edge of the image, and outputs a third image. The edge enhancement unit includes a high-pass filter for receiving the third image to enhance the edge of the image.

In order to have a better understanding of the above and other aspects, some embodiments are specifically cited below and described in detail with reference to the accompanying drawings.

Description of drawings

FIG. 1 is a schematic diagram illustrating a digital image.

FIG. 2 illustrates a cross-shaped pattern centered on a target pixel in a region of a digital image.

FIG. 3 is a flowchart of an embodiment of an image edge processing method.

FIG. 4 is a block diagram of an embodiment of an image processing device.

FIG. 5 is a block diagram of an embodiment of the edge low-pass filtering unit in FIG. 4 .

FIG. 6 is a block diagram of an embodiment of an electronic device capable of implementing the image edge processing method in FIG. 3 .

7 and 8 are flowcharts of other embodiments of the image edge processing method.

Description of reference symbols

1: Digital image

10: An area of a digital image

40: Image processing device

50, 410: noise suppression unit

60: Electronics

101, 102: cross pattern

420: edge low-pass filter unit

430: Edge Strengthening Unit

510: The first arithmetic circuit

520: Second operation circuit

521: judgment circuit

525: Estimation circuit

610: processing unit

620: storage unit

630: Image module

640: display unit

S70, S80, S100: Steps

S110, S120, S125, S130, S135, S139: steps

S810, S821, S822: steps

Detailed ways

Embodiments of an image edge processing method and an image processing device are proposed below. FIG. 3 is a flowchart of an embodiment of an image edge processing method. This embodiment can be used to process a region 10 centered on a target pixel PC in an input image 1 (which may be a static image or a dynamic image) shown in FIG. 1 to generate an estimated pixel value of the target pixel PC. Applying this embodiment, each pixel in the input image 1 is used as the target pixel. If this method is used in a certain order, such as row by row, and the estimated pixel value is used to replace the corresponding target pixel, an edge-processed image can finally be generated. An output image. In some embodiments, compared with the input image, the edge-processed output image can obtain an edge-smoothing effect. This method can be implemented in various image processing devices, such as the edge low-pass filter unit 420 of the image processing device 40 (such as an image processing chip or image trigger) in Figure 4 to implement this method, and the electronic device in Figure 6 A vision module or processing unit in implements the method.

Please refer to FIG. 3, as shown in step S110, in an area of an input image, for a cross-shaped pattern centered on a target pixel in this area, a first direction gradient along a first direction and a gradient along a first direction are calculated. A second direction gradient of the second direction. For example, FIG. 2 shows a cross-shaped pattern 101 or 102 centered on a target pixel PC in an area 10 of an input image 1 . In addition, the first direction and the second direction are, for example, the horizontal direction and the vertical direction, respectively, and vice versa.

As shown in step S120, it is determined whether the gradient in the first direction is greater than a first comparison gradient value, wherein the first comparison gradient value is determined according to the gradient in the second direction. If the judgment result of step S120 is yes, as shown in step S125, based on the pixel values of the first plurality of pixels in the cross-shaped pattern along the second direction, a first estimated pixel value is obtained as the target pixel. An estimated pixel value. If the judgment result of step S120 is negative, as shown in step S130, it is judged whether the second direction gradient is greater than a second comparison gradient value, wherein the second comparison gradient value is determined according to the first direction gradient. If the judgment result of step S130 is yes, as shown in step S135, based on the pixel values of the second plurality of pixels in the cross-shaped pattern along the first direction, a second estimated pixel value is obtained as the target pixel. Estimated pixel values. If the determination result of step S130 is negative, then as shown in step S139 , output the pixel value of the target pixel (ie regarded as estimated pixel value).

Various implementations of the steps of the method in FIG. 3 will be described below by taking the cross patterns 101 and 102 as examples.

For step S110 , one to a plurality of unit gradients in the first direction along a first direction (eg, vertical direction) and at different distances from the target pixel PC in the cross pattern are calculated. If calculated by difference, the gradient of one or more first direction units of the cross-shaped pattern 101 can be expressed as: |PC-V1|, |PC-V2|, where as shown in Figure 2, V1 and V2 respectively represent the cross-shaped pattern The pixel value of two pixels in the vertical direction within 101. One or more first direction unit gradients of the cross-shaped pattern 102 can be expressed as: |PC-V1|, |PC-V2|, |V0-V1|, |V2-V3|, wherein V0, V1 as shown in FIG. 2 , V2 and V3 respectively represent the pixel values of 4 pixels in the vertical direction in the cross pattern 101 .

The first direction gradient, for example, the sum of gradients of one or more first direction units. For example, each of the one or more unit gradients in the first direction is the sum of absolute differences between two groups of adjacent two pixels located on two different sides of the target pixel along the first direction in the cross pattern. For the cross-shaped pattern 101 , the gradient in the first direction is written as deltaV=|PC-V1|+|PC-V2|. For the cross-shaped pattern 102, the gradient in the first direction is written as deltaV=|PC-V1|+|PC-V2|+|V0-V1|+|V2-V3|.

One or more unit gradients in the second direction along a second direction (such as a horizontal direction) and different distances from the target pixel PC in the cross pattern 101 are calculated. If calculated by difference, the gradient of one or more first direction units of the cross-shaped pattern 101 can be expressed as: |PC-H1|, |PC-H2|, where as shown in Figure 2, H1 and H2 respectively represent the cross-shaped pattern The pixel value of two pixels in the horizontal direction within 101. One or more unit gradients in the first direction of the cross-shaped pattern 102 can be expressed as: |PC-H1|, |PC-H2|, |H0-H1|, |H2-H3|, where H0, H1 , H2 and H3 respectively represent pixel values of four pixels in the horizontal direction in the cross-shaped pattern 101 .

The second direction gradient is, for example, the sum of the one or more second direction unit gradients. For example, each of the one or more unit gradients in the second direction is the sum of the absolute differences between two groups of adjacent two pixels located on two different sides of the target pixel along the second direction in the cross pattern. For the cross-shaped pattern 101 , the gradient in the second direction is expressed as: deltaH=|PC-H1|+|PC-H2|.

For step S120, for example, the first comparative gradient value is substantially equal to the product of the second directional gradient and a predetermined multiple (such as expressed as W_Edge), following the above example, the first comparative gradient value can be expressed as: W_Edge x deltaH. The second comparison gradient value, for example, is substantially equal to the product of the gradient in the first direction and the predetermined multiple, such as: W_Edge x deltaV.

In step S125, the first estimated pixel value, for example, is substantially equal to a weighted sum of pixel values of the first plurality of pixels. In step S135, the second estimated pixel value is, for example, substantially equal to a weighted sum of the pixel values of the second plurality of pixels. Using the above example, taking the cross-shaped pattern 101 as an example, the first estimated pixel value can be expressed as: the weighted sum of H1 and H2, and the second estimated pixel value can be expressed as: the weighted sum of V1 and V2, wherein each pixel value The corresponding weighted value (or weight) may be set to 1/2, for example. Taking the cross pattern 102 as an example, the first estimated pixel value can be expressed as: the weighted sum of H0, H1, H2 and H3, and the second estimated pixel value can be expressed as: the weighted sum of V0, V1, V2 and V3, The weight can be set to 1/2, for example.

In addition, in some embodiments of steps S125 and S135, when calculating the first estimated pixel value (or the second estimated pixel value) according to the above example, the first plurality of pixels (or the second plurality of pixels) respectively correspond to The weighted value is determined according to the relative distances between the first plurality of pixels (or the second plurality of pixels) and the target pixel. For example, when the relative distance is larger, the corresponding weighting value is smaller. Take the cross-shaped pattern 102 as an example, the first estimated pixel value can be expressed as: w0*H0+w1*H1+w2*H2+w3*H3, wherein because the two pixels of the pixel values H1 and H2 are different from the target pixel PC The relative distance is less than two pixels from the pixel values H0 and H3, so the weighted values w0 and w3 can be set to be smaller than the weighted values w1 and w2 respectively. In accordance with this example related to the relative distance of the target pixel, for example, it can be set that the first estimated pixel value is:

(W_P0*(H1+H2)+W_P1*(H0+H3))/(W_P0+W_P1),

Where W_P0 and W_P1 are real values and W_P0 is greater than W_P1. Relatively speaking, for the calculation of the second estimated pixel value, the above-mentioned example of the first estimated pixel value can also be followed to set the size of the weighted value calculated based on the weighted sum of the second plurality of pixels, for example, it can be set, The second estimated pixel value is:

(W_P0*(V1+V2)+W_P1*(V0+V3))/(W_P0+W_P1).

The above-mentioned embodiments of calculating estimated pixel values with weights can be regarded as a weighted spatial filter (Spatial Filter) because the weights are given according to the position distance.

In addition, the significance of steps S120 to S125 is that when the difference between the value of the target pixel and adjacent pixels in the first direction (such as the vertical direction) in the cross-shaped pattern reaches a certain level, the target pixel may be affected by noise or appear uneven In the case of an edge, the target pixel is compensated by adjacent pixel values in the second direction (such as the horizontal direction) with relatively small (or smoother or less affected by noise) relative to the pixel value drop, that is, an estimated pixel value is obtained. Furthermore, the meanings of steps S130 to S135 also have similar functions, so details are not repeated here. The meaning of step S139 is that when the condition is satisfied, no compensation is needed, so the pixel value of the target pixel is output. In this way, if the weighted sum is used and the relationship between the weighted value and the relative distance is considered to obtain the estimated pixel value as in some embodiments above, an output image after edge processing will be able to obtain a smoother edge quality effect.

In addition, in the above-mentioned embodiment, the first estimated pixel value and the second pixel value have nothing to do with the pixel value of the target pixel. In other embodiments, the target pixel can be included in the calculation of the first estimated pixel value or the second pixel value, but the weighted value (or weight) of the target pixel is preferably set to be smaller than the weighted values of other pixel values, so as to Reduce the effect of noise.

7 and 8 are flowcharts of other embodiments of the image edge processing method. In FIG. 7 , as shown in step S70 , one of multiple filters is selected, wherein these filters respectively correspond to cross patterns of different sizes. For example, provide at least two or more cross patterns, such as 101 and 102, and select one of them. Next, as represented by step S100 : perform the steps shown in FIG. 3 according to the cross-shaped pattern corresponding to the selected filter. In addition, the execution order of the steps is not limited, as long as the estimated pixel values can be obtained; for example, step S110 may calculate gradients for these filters, then execute step S70, and then execute other steps. In some other embodiments, one of the filters is selected according to an analog gain value. For example, in FIG. 8, step S80 represents an implementation example of step S70. As shown in step S810, it is determined whether an analog gain value is greater than or equal to a threshold value, such as 4. If yes, as shown in step S822, a second filter is set as the selected filter, wherein the second filter corresponds to a larger cross pattern, such as 102 in FIG. 2 . If not, as shown in step S821, a first filter is set as the selected filter, wherein the first filter corresponds to a smaller cross-shaped pattern, such as 101 in FIG. 2 . Here, because the larger the analog gain value, it means that the noise of the input image may be amplified, so a larger cross-shaped filter is selected to obtain a smoother effect.

In some embodiments, since the input image may be based on an analog gain value as the analog gain or the result of selecting a larger ISO value (such as 400, 800 or above) or the result of shooting when the light sensitivity is insufficient, this situation can be based on FIG. 3 The method uses a cross-shaped filter, or selects one of a plurality of cross-shaped filters of different sizes to obtain an estimated pixel value, thereby obtaining a processed output image, and Carry out the next step of image processing (such as edge enhancement processing). In addition, the cross-shaped pattern can be, for example, 3x3, 5x5, 7x7 or other cross-shaped patterns with different lengths and widths, and the cross-shaped patterns are all within an area centered on a target pixel. Therefore, the area 10 shown in FIG. 1 can also be changed to 3x3, 5x5, 7x7 pixels or other areas with different lengths and widths.

In addition, the surrounding pixels of the input image, each pixel on the right and above (that is, the first and second rows or columns) of a target pixel PC as shown in Figure 1, can also be processed using the method shown in Figure 3 or other methods deal with. When these surrounding pixels are used as a new target pixel, it may not be possible to obtain an area as large as the area 10 from the input image. However, point supplementation can be used. For example, when surrounding pixels are used as new target pixels, the target pixels themselves are supplemented to obtain an area with the same size as the area 10, and the processing is performed in the manner shown in FIG. 3 . In other embodiments, when the pixels in the second row or column are used as target pixels, a smaller area (such as 3×3) may be set; for the pixels in the first row or column, no processing may be performed. There are many ways to implement these methods, so they are not limited to the above.

In addition, various embodiments of the above methods can be implemented in various image processing devices, such as the edge low-pass filter unit 420 of the image processing device 40 (such as an image processing chip, image trigger or image sensor) shown in FIG. 4 The method is implemented, and the image module or processing unit in the electronic device as shown in FIG. 6 implements the method.

FIG. 4 is a block diagram of an embodiment of an image processing device.

The image processing device 40 includes: a noise suppression unit 410 , an edge low-pass filter unit 420 and an edge enhancement unit 430 . The noise suppression unit 410 includes a low-pass filter, receives a first image, and performs a noise suppression process to output a second image. The edge low-pass filtering unit 420 receives the second image (ie the input image described in the method of FIG. 3 ), performs an edge low-pass filtering process to smooth the edge of the image, and outputs a third image. The edge enhancement unit 430, which includes a high-pass filter, receives the third image to enhance the edge of the image.

The edge low-pass filtering unit 420 performs the following operations: For a cross-shaped pattern centered on a target pixel in the second image, calculate a first direction gradient along a first direction and a second direction gradient along a second direction direction gradient; and according to the first direction gradient and the second direction gradient, determine whether to compensate the target pixel based on the pixel values of the first plurality of pixels along the second direction in the cross-shaped pattern, or based on the pixel values along the first direction The pixel value of the second plurality of pixels is used to compensate the target pixel, or output the pixel value of the target pixel. That is to say, the edge low-pass filtering unit 420 can implement its circuit hardware according to steps S110 to S139 as shown in FIG. 3 . In one embodiment, the edge low-pass filter unit 420 is coupled between the noise suppression unit 410 and the edge enhancement unit 430 and operates in a pipeline manner.

FIG. 5 is a block diagram of an embodiment of the edge low-pass filtering unit in FIG. 4 . The edge low-pass filter unit 50 includes a first operation circuit 510 and a second operation circuit 520 . The first arithmetic circuit 510 has a circuit for realizing step S110. The second arithmetic circuit 520 has a circuit that realizes steps S120 to S139. The second computing circuit 520 includes a judging circuit 521 and an estimating circuit 525 . The judgment circuit 521 has a circuit that realizes steps S120 and S130. The estimation circuit 525 has a circuit for realizing steps S125, S135, and S139.

FIG. 6 is a block diagram of an embodiment of an electronic device capable of implementing the image edge processing method in FIG. 3 . The electronic device 60 is, for example, various devices capable of image processing, such as digital cameras, smart phones, tablet computers, multimedia devices, televisions or computers. The electronic device 60 includes a processing unit 610 , a storage unit 620 , an image module 630 and a display unit 640 . In an example, the image module 630 may include the image processing device 40 as shown in FIG. 4 . In another example, the processing unit 610 reads program instructions from the storage unit 620 to implement the embodiment of the method according to FIG. 3 .

As shown in the examples shown in FIGS. 4 to 6 above, the embodiment of the image edge processing method can use integrated circuits such as microcontrollers, microprocessors, digital signal processors, and application specific integrated circuits (ASIC, Application Specific Integrated Circuit) Or element programmable logic gate array (FPGA, Field Programmable Gate Array) or a logic circuit to implement.

In the above-mentioned embodiments, each pixel may represent R/G/B values or Y (brightness) or any different color coordinate conversion values. For each target pixel of an input image, the values corresponding to different components (component) or subpixels (subpixel) of the target pixel, such as Y or R, can be individually or separately obtained by applying the embodiment of the method in FIG. 3 corresponding to the subpixel The estimated pixel value of the value.

The above embodiments provide an image edge processing method and an image processing device. In some embodiments, due to the gradient judgment, the resolution can be preserved, and for example, smoothing is done on the vertical or horizontal edge, so that the edge part can be smoothed (smooth repair), and high-frequency noise (Noises and Artifacts) can be reduced. Make the image look clearer. In addition, in some embodiments, it is mentioned that some coefficients are used to control the edge processing (such as the predetermined multiple value in step S120 or S130 ) to determine whether to perform calculations for estimating pixel values, which can effectively reduce the required memory space. In addition, some embodiments may perform different degrees of processing according to different filters, so that the implementation of the method has greater flexibility. When this method is implemented, the computational complexity is not high, and it is easy to integrate into various image processing devices, platforms, and software. Furthermore, according to the quantification method, its effect can be verified experimentally.

To sum up, although the present invention has been disclosed by the above embodiments, it is not intended to limit the implementation of the present invention. Those skilled in the art can make various changes and modifications under the premise of following the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be determined by the claims of the present invention.

Claims (23)

1. An image edge processing method, comprising: (a) In a region of an input image, for a cross-shaped pattern centered at a target pixel in the region, calculate a first direction gradient along a first direction and a second direction along a second direction gradient; (b) judging whether the gradient in the first direction is greater than a first comparison gradient value, wherein the first comparison gradient value is determined according to the gradient in the second direction; (c) if the determination result of step (b) is yes, then based on the pixel values of the first plurality of pixels in the cross-shaped pattern along the second direction, a first estimated pixel value is obtained as the target pixel an estimated pixel value; (d) If the judgment result of step (b) is no, then judge whether the second direction gradient is greater than a second comparison gradient value, wherein the second comparison gradient value is determined according to the first direction gradient; (e) if the determination result of step (d) is yes, then based on the pixel values of the second plurality of pixels in the cross-shaped pattern along the first direction, a second estimated pixel value is obtained as the target pixel the estimated pixel value; (f) If the judgment result of step (d) is no, then output the pixel value of the target pixel. 2. The edge processing method of image as claimed in claim 1, further comprising: calculating gradients of one or more first direction units along a first direction and at different distances from the target pixel in the cross pattern; calculating gradients of one to a plurality of second direction unit gradients along a second direction and different distances from the target pixel in the cross pattern, and wherein In step (a), the first directional gradient is the sum of the one or more first directional unit gradients, and The second directional gradient is the sum of the one or more second directional unit gradients. 3. The edge processing method of image as claimed in claim 2, wherein Each of the one or more unit gradients in the first direction is the sum of the absolute differences between two groups of adjacent two pixels located on two different sides of the target pixel along the first direction in the cross pattern, and Each of the one or more unit gradients in the second direction is a sum of absolute differences between two groups of adjacent two pixels located on two different sides of the target pixel along the second direction in the cross pattern. 4. The edge processing method of image as claimed in claim 1, wherein The first comparison gradient value is substantially equal to the product of the second direction gradient and a predetermined multiple, and The second comparison gradient value is substantially equal to the product of the first direction gradient and the predetermined multiple. 5. The image edge processing method according to claim 1, wherein the first estimated pixel value is substantially equal to a weighted sum of pixel values of the first plurality of pixels; and The second estimated pixel value is substantially equal to a weighted sum of pixel values of the second plurality of pixels. 6. The image edge processing method according to claim 5, wherein the weighted values respectively corresponding to the first plurality of pixels and the second plurality of pixels are based on the respective weight values of the first plurality of pixels and the second plurality of pixels The relative distance to the target pixel is determined. 7. The image edge processing method according to claim 6, wherein when the relative distance is larger, the corresponding weighting value is smaller. 8. The edge processing method of image as claimed in claim 3, further comprising: Select one of a plurality of filters, wherein the filters correspond to cross-shaped patterns of different sizes, and wherein Steps (a)-(f) are performed according to the cross pattern corresponding to the selected filter. 9. The image edge processing method as claimed in claim 8, wherein the step of selecting one of the filters is determined according to an analog gain value. 10. The image edge processing method as claimed in claim 1, wherein the first estimated pixel value is independent of the second pixel value and the target pixel value. 11. An image processing device, comprising: A noise suppression unit, which includes a low-pass filter, and receives a first image, performs a noise suppression process to output a second image; An edge low-pass filter unit receives the second image, performs an edge low-pass filter process to smooth the edge of the image, and outputs a third image; and An edge enhancement unit, including a high-pass filter, receives the third image to enhance the edge of the image. 12. The image processing device as claimed in claim 11, wherein the edge low-pass filter unit performs the following operations: calculating a first directional gradient along a first direction and a second directional gradient along a second direction for a cross-shaped pattern centered on a target pixel in the second image; and According to the gradient in the first direction and the gradient in the second direction, it is determined whether to compensate the target pixel based on the pixel values of the first plurality of pixels in the cross pattern along the second direction, or to compensate the target pixel based on the pixel values along the first direction. The pixel value of the second plurality of pixels is used to compensate the target pixel, or output the pixel value of the target pixel. 13. The image processing device as claimed in claim 12, wherein the edge low-pass filter unit performs the following operations: judging whether the first direction gradient is greater than a first comparison gradient value, wherein the first comparison gradient value is determined according to the second direction gradient; and When the determination result is yes, a first estimated pixel value is obtained as an estimated pixel value of the target pixel based on the pixel values of the first plurality of pixels. 14. The image processing device according to claim 13, wherein if the result of the determination is negative, the edge low-pass filter unit performs the following operations: Then judge whether the second direction gradient is greater than a second comparison gradient value, wherein the second comparison gradient value is determined according to the first direction gradient; and When the result of the re-judgment is yes, a second estimated pixel value is obtained as an estimated pixel value of the target pixel based on the pixel values of the second plurality of pixels. 15. The image processing device according to claim 14, wherein if the result of the re-judgment is negative, the edge low-pass filter unit outputs the pixel value of the target pixel. 16. The image processing device as claimed in claim 13, wherein the first comparison gradient value is substantially equal to the product of the second direction gradient and a predetermined multiple. 17. The image processing device according to claim 14, wherein the second comparison gradient value is substantially equal to the product of the first direction gradient and a predetermined multiple. 18. The image processing device of claim 13, wherein the first estimated pixel value is substantially equal to a weighted sum of pixel values of the first plurality of pixels. 19. The image processing device of claim 14, wherein the second estimated pixel value is substantially equal to a weighted sum of pixel values of the second plurality of pixels. 20. The image processing device according to claim 11, wherein the edge low-pass filter unit comprises: a first computing circuit, for a cross-shaped pattern centered on a target pixel in the second image, calculating a first directional gradient along a first direction and a second directional gradient along a second direction; and A second operation circuit, according to the gradient in the first direction and the gradient in the second direction, determine whether to compensate the target pixel based on the pixel values of the first plurality of pixels along the second direction in the cross-shaped pattern, or based on Compensating the target pixel with pixel values of a second plurality of pixels along the first direction, or outputting the pixel value of the target pixel. 21. The image processing device as claimed in claim 20, wherein the second computing circuit judges whether the first direction gradient is greater than a first comparison gradient value, wherein the first comparison gradient value is determined according to the second direction gradient, Wherein when the determination result is yes, the second operation circuit outputs a first estimated pixel value as an estimated pixel value of the target pixel based on the pixel values of the first plurality of pixels. 22. The image processing device as claimed in claim 21, wherein if the result of the determination is negative, the second operation circuit then determines whether the second direction gradient is greater than a second comparison gradient value, wherein the second comparison The gradient value is determined according to the gradient in the first direction, and when the result of the re-judgment is yes, the second operation circuit outputs a second estimated pixel value as the target pixel based on the pixel values of the second plurality of pixels An estimated pixel value of . 23. The image processing device according to claim 22, wherein if the result of the re-judgment is negative, the second operation circuit outputs the pixel value of the target pixel.

Images