An image processing method for uniformizing pixel size

Abstract

In order to solve the problem of inconsistent image pixel size in human eye tracking 3D display device, the application provides an image processing method for pixel size uniformization. The method obtains a final display image through image resolution preprocessing, and compensates the display size of the pixel through special pixel arrangement, so that each pixel can be displayed with the same physical size after the column lens. Specifically, each pixel column in the preprocessed parallax image will be mapped to two pixel columns in the parallax synthesis image, and the two pixel columns will be displayed continuously at the rear and front of the adjacent column lens, so that each preprocessed parallax image pixel column has the same total duty cycle, and finally the pixel size uniformization is realized.

Description

Technical Field

[0001] This invention belongs to the field of display technology, and more specifically, this invention relates to an image processing method for uniformizing pixel size. Background Technology

[0002] Existing lenticular 3D displays typically consist of a 2D display panel coupled with a lenticular lens grating, and use eye tracking to display stereoscopic images. The display dynamically adjusts the pixel distribution on the display panel according to the viewer's eye position, thereby projecting the correct parallax image onto the viewpoint. To address crosstalk issues under dynamically changing eye positions, the viewpoint is usually positioned to see multiple columns of pixels through a single lenticular lens grating, and these columns are set to originate from the same parallax image. However, under random eye positions, aside from specific locations, it is more common for these seen pixels to not be displayed at the same scale; that is, the left and right pixel columns are displayed in different proportions. Consequently, traditional techniques result in non-uniform pixel sizes, leading to degraded display quality. To solve this problem, this invention proposes an image processing method for pixel size uniformity. This method obtains the final display image through image resolution preprocessing and compensates for the pixel display size through a special pixel arrangement, ensuring that each pixel is displayed at the same physical size after passing through the lenticular lens. Summary of the Invention

[0003] To address the issue of inconsistent image pixel sizes in eye-tracking 3D display devices, this invention proposes an image processing method for achieving pixel size uniformity.

[0004] This image processing method for pixel size uniformity includes the following steps: (1) Image resolution preprocessing Preprocess all N original disparity images. Let the resolution of the original disparity images be X×Y (X rows, Y columns of pixels). Then, transform the resolution of each original disparity image into a preprocessed disparity image of resolution J×K, where J and K satisfy the following conditions: J=O; K = P / 2N + 1; Where O is the number of pixel rows in the parallax composite image, P is the number of pixel columns in the parallax composite image, and N is the number of viewpoints on the 3D display. (2) Pixel arrangement processing For the i-th column of the n-th preprocessed disparity image, the pixels are placed in the j-th and k-th columns of the disparity composite image, where j = (i-2)×2N+2n and k = (i-1)×2N+2n-1, respectively, and 1≤j≤P, 1≤k≤P and n∈{1,2,3,…,N}.

[0005] Optionally, the edge columns of the parallax composite image are set to 0.

[0006] The working principle of this invention is as follows: The N-viewpoint 3D display consists of a 2D display panel and a lenticular lens grating, placed sequentially from back to front. To reduce crosstalk, the distance d from the lenticular lens grating to the 2D display panel is less than the focal length f of the lenticular lens grating, so that at the optimal viewing distance, the human eye can see two adjacent columns of pixels through a single lenticular lens. The resolution of the parallax composite image on the 2D display panel is 0×P.

[0007] The original N disparity images each have a resolution of X×Y. To ensure that their pixel values ​​are correctly filled into the O×P array, they should be preprocessed to convert them into a preprocessed disparity image with a resolution of J×K. Here, J=O, meaning that the number of pixel rows in the preprocessed disparity image is the same as the number of rows in the composite disparity image; and K=P / 2N+1, which ensures that it matches the number of columns required for the composite disparity image in subsequent processing.

[0008] Furthermore, the pixel arrangement is processed. The pixels in the i-th column of the n-th preprocessed disparity image are placed in the j-th and k-th columns of the disparity composite image, where j = (i-2)×2N+2n and k = (i-1)×2N+2n-1, and 1 ≤ j ≤ P and 1 ≤ k ≤ P.

[0009] From j=(i-2)×2N+2n; k=(i-1)×2N+2n-1, we know that for any preprocessed disparity image, j and k are arranged according to i in 2N periods, that is, there are 2 columns of pixels belonging to the same disparity image within each 2N period; and kj=2N-1, we know that the 2 columns of pixels belonging to the same disparity image must be adjacent.

[0010] Specifically, the (i-1)th column of pixels k in the nth preprocessed disparity image i-1 =(i-2)×2N+2n-1, while the j-th column of pixels in the nth preprocessed disparity image has j i =(i-2)×2N+2n, at this time j i -k i-1 =1, meaning that pixels in adjacent columns in the preprocessed disparity image are still placed adjacent to each other in the composite disparity image, i.e., the (i-1)th column is adjacent to the ith column, and the ith column is after the (i-1)th column. Similarly, for the i-th column of pixels k in the n-th preprocessed disparity image... i =(i-1)×2N+2n-1 and the (i+1)th column pixel j in the nth preprocessed disparity image i+1 =(i-1)×2N+2n, at this time j i+1 -k i=1, meaning that the i-th column and the (i+1)-th column are also adjacent, and the (i+1)-th column is after the i-th column; Considering that there are two columns of pixels belonging to the same disparity image within each 2N period, the two columns of pixels in the previous period are the (i-1)th column and the ith column respectively, and the two columns of pixels in the next period are the ith column and the (i+1)th column respectively. At this time, for any column of pixel i, it exists simultaneously in the rear of the two columns of pixels in the previous period and in the front of the two columns of pixels in the next period.

[0011] Because the light is split by the cylindrical lens grating, the same cylindrical lens will see two columns of pixels, with the duty cycle of the left column of pixels being u and the duty cycle of the right column of pixels being v, so u+v=1.

[0012] Since any pixel i in a given column exists simultaneously at the rear of two columns of pixels in the previous period and at the front of two columns of pixels in the next period, its proportion in the front cylindrical lens is v, and its proportion in the rear cylindrical lens is u. The sum of these two proportions is a constant 1. Therefore, regardless of the duty cycles u and v of the left and right pixels, this invention ensures that any pixel in the i-th column has the exact same total duty cycle, i.e., the exact same pixel size. Ultimately, this invention achieves pixel size uniformity by displaying the pre-processed parallax image pixel columns across cylindrical lenses. Attached Figure Description

[0013] Figure 1 This is a schematic diagram illustrating the image resolution preprocessing principle of the present invention.

[0014] Figure 2 This is a schematic diagram illustrating the pixel arrangement processing principle of the present invention.

[0015] Figure 3 This is a schematic diagram of the optical structure adapted to the present invention.

[0016] Figure 4 This is a schematic diagram illustrating the technical principle of the present invention.

[0017] Figure 5 This is a schematic diagram of the effect of a traditional structure.

[0018] Figure 6 This is a schematic diagram of the edge column setting 0 of the present invention.

[0019] Figure 7 This is a schematic diagram illustrating the pixel arrangement processing principle of setting 0 at the edge column in this invention.

[0020] Figure 8 This is a schematic diagram illustrating the technical principle of setting 0 at the edge of the present invention.

[0021] Icons: 100 - Original parallax image, 200 - Preprocessed parallax image, 300 - Parallax composite image, 400 - Cylindrical grating.

[0022] It should be understood that the above figures are only schematic and are not drawn to scale. Detailed Implementation

[0023] Figure 1 This is a first embodiment of an image processing method for pixel size uniformity.

[0024] This image processing method for pixel size uniformity includes the following steps: (1) Image resolution preprocessing Please refer to Figure 1 Preprocess the four original disparity images 100 to be synthesized. The resolution of the original disparity images is X×Y, where X=2560 and Y=3840 (2560 rows and 3840 columns of pixels). Then, convert their resolution to a preprocessed disparity image 200 with a resolution of J×K, where J and K satisfy the following conditions: J=O; K = P / 2N + 1; Where O=1440 is the number of pixel rows in the parallax composite image, P=2560 is the number of pixel columns in the parallax composite image, and N=4 is the number of viewpoints on the 3D display. Specifically, J equals 1440, and K=321.

[0025] (2) Pixel arrangement processing The i-th column of pixels in the n-th preprocessed disparity image is placed in the j-th and k-th columns of the disparity composite image 300, where j = (i-2)×2N+2n and k = (i-1)×2N+2n-1, and 1 ≤ j ≤ P, 1 ≤ k ≤ P, and n ∈ {1,2,3,…,N}.

[0026] Please refer to Figure 2 Taking the first preprocessed disparity image as an example, in each element axy, x represents the row number and y represents the column number. a12 is the pixel in the 1st row and 2nd column, and similarly, a34 represents the pixel in the 3rd row and 4th column. Now, taking the 3rd column pixel as an example, with n=1 and i=3, the 3rd pixel column, i.e., all ax3 (x refers to any value from 1 to J), is placed in the j=10th and k=17th columns of the disparity composite image 300. Other pixel columns of this preprocessed disparity image are filled according to the same rule. For example, the 2nd column pixel is placed in the j=2nd and k=9th columns of the disparity composite image 300. Similarly, the 2nd column pixel of the 2nd preprocessed disparity image is placed in the j=4th and k=11th columns. They all satisfy j=(i-2)×2N+2n; k=(i-1)×2N+2n-1.

[0027] The working principle of this embodiment is as follows: Please refer to Figure 3The 4-viewpoint 3D display consists of a 2D display panel and a lenticular lens grating 400. The 2D display panel is used to display the parallax composite image 300. The 2D display panel and the lenticular lens grating are placed sequentially from back to front. To reduce crosstalk, the distance from the lenticular lens grating to the 2D display panel is d, where d = 4.2 mm, which is less than the focal length of the lenticular lens grating, f, where f = 5 mm. At this optimal viewing distance, the human eye can see an upright virtual image of the pixel, and the lenticular lens pitch is S, where S = 1.27 mm, and the pixel size is T, where T = 0.1554 mm. Therefore, through one lenticular lens, two adjacent columns of pixels can always be seen. The resolution of the parallax composite image on the 2D display panel is 1440 × 2560.

[0028] The original N=4 disparity images each have a resolution of 2560×3840 (X×Y). To ensure that their pixel values ​​are correctly filled into a 1440×2560 (O×P) array, they should undergo resolution preprocessing, transforming them into a preprocessed disparity image with a resolution of J×K (1440×321). Here, J=O, meaning the number of rows in the preprocessed disparity image is the same as the number of rows in the composite disparity image; and K=P / 2N+1, ensuring that it matches the number of columns required for the composite disparity image in subsequent processing.

[0029] Furthermore, the pixel arrangement is processed. The pixels in the i-th column of the n-th preprocessed disparity image are placed in the j-th and k-th columns of the disparity composite image, where j = (i-2)×2N+2n and k = (i-1)×2N+2n-1, and 1 ≤ j ≤ P and 1 ≤ k ≤ P.

[0030] Since N=4, the above formula can be expressed as j= (i-2)×8+2n; k=(i-1)×8+2n-1; and 1≤j≤P and 1≤k≤P can make the values ​​of j and k take values ​​between 1 and 2560 without overflow.

[0031] From j=(i-2)×8+2n; k=(i-1)×8+2n-1, we know that for any preprocessed disparity image, j and k are arranged according to i with a period of 8 columns of pixels. That is, within one period of 8 columns of pixels, there are 2 columns of pixels belonging to the same disparity image. And kj=2N-1, we know that the 2 columns of pixels belonging to the same disparity image must be adjacent.

[0032] Specifically, the (i-1)th column of pixels k in the nth preprocessed disparity image i-1 =(i-2)×2N+2n-1, while the j-th column of pixels in the nth preprocessed disparity image has j i =(i-2)×2N+2n, at this time j i -k i-1=1, meaning that pixels in adjacent columns in the preprocessed disparity image are still placed adjacent to each other in the composite disparity image, i.e., the (i-1)th column is adjacent to the ith column, and the ith column is after the (i-1)th column. Please refer to Figure 2 Taking the third column of the first preprocessed parallax image as an example, its j3=10 and k3=17, while the second column of the first preprocessed parallax image has j2=2 and k2=9. Therefore, it can be seen that j... i -k i-1 =1.

[0033] Similarly, for the i-th column of pixels k in the n-th preprocessed disparity image... i =(i-1)×2N+2n-1 and the (i+1)th column pixel j in the nth preprocessed disparity image i+1 =(i-1)×2N+2n, at this time j i+1 -k i =1, meaning that the i-th column and the (i+1)-th column are also adjacent, and the (i+1)-th column is after the i-th column; Please refer to Figure 2 Taking the third column of the first preprocessed parallax image as an example, its j3=10 and k3=17, while the fourth column of the first preprocessed parallax image has j4=18 and k4=25. Therefore, it can be seen that j... i+1 -k i =1.

[0034] Considering that there are two columns of pixels belonging to the same disparity image within each 2N period, the two columns of pixels in the previous period are the (i-1)th column and the ith column respectively, and the two columns of pixels in the next period are the ith column and the (i+1)th column respectively. At this time, for any column of pixel i, it exists at the rear of the two columns of pixels in the previous period and at the front of the two columns of pixels in the next period. Please refer to Figure 2 Taking the third column of the first preprocessed parallax image as an example, namely ax3, it is located after the two columns of pixels in the second period, that is, after ax2, and in front of the two columns of pixels in the third period, that is, before ax4.

[0035] Please refer to Figure 4 Because after the beam is split by the cylindrical lens grating 400, the same cylindrical lens will see two columns of pixels, and the duty cycle of the left column of pixels is u, and the duty cycle of the right column of pixels is v, so u+v=1. Specifically, at the current viewing position, the width of the left column of pixels in the cylindrical lens is W1=0.8 mm, and the width of the right column of pixels in the cylindrical lens is W2=0.47 mm. Their sum is the cylindrical lens pitch S=1.27 mm. Therefore, u= W1 / S=0.63, v=W2 / S=0.37, and u+v=1.

[0036] Since any pixel i exists simultaneously in the rear of two columns of pixels in the previous period and in the front of two columns of pixels in the next period, its proportion in the front cylindrical lens is v, and its proportion in the rear cylindrical lens is u. The sum of the two is exactly a constant 1. That is, regardless of the duty cycles u and v of the left and right side pixels, the present invention can make any pixel in the i-th column have the same total duty cycle, that is, the same pixel size. Please refer to Figure 4 ,right Figure 4 All pixels from ax2 to ax(K-1) are displayed within two adjacent cylindrical lenses, and each pixel has a width of S = 1.27 mm. Ultimately, this embodiment achieves pixel size uniformity by displaying the preprocessed parallax image pixels across cylindrical lenses.

[0037] Please refer to Figure 5 , Figure 5 The technical effect of traditional parallax composite images is demonstrated. The pixel column of any parallax image is distributed only once in the parallax composite image. Therefore, it will inevitably lead to the display ratio of pixels in odd-numbered columns being greater than that in even-numbered columns, resulting in uneven pixel display.

[0038] Furthermore, in Figure 4 In this embodiment, ax1 occupies only the first cylindrical lens, and it can be foreseen that axK will only occupy the rear part of the last Kth column of cylindrical lenses. Their pixel widths are W1 = 0.8 mm and W2 = 0.47 mm, respectively. Therefore, a second embodiment of the present invention is proposed, which can be found in the following reference. Figure 6 .

[0039] Figure 6 In the preprocessed parallax image 200, the first and last columns are set to 0; at this point, please refer to... Figure 7 When performing pixel arrangement processing, the first part of the first cycle is filled with all 0 values ​​of ax1, and similarly, the last part of the last cycle is filled with all 0 values ​​of axK.

[0040] Please refer to Figure 8 Finally, after it is displayed through the lenticular lens grating, the front part of the first lenticular lens will be completely black, and the rear part of the last lenticular lens will also be completely black. At this time, the pixel width of all the remaining pixels is S=1.27 mm.

[0041] Please refer to Figure 2 In the third embodiment of the present invention, the cylindrical lens pitch S=0.99 mm and 2N=6.375, which can realize up to 3-viewpoint stereoscopic display.

[0042] In the third embodiment, since 2N=6.375 is not an integer, j=(i-2)×2N+2n and k=(i-1)×2N+2n-1 are rounded to the nearest integer and arranged so that the width of each pixel is S=0.99 mm.

[0043] Ultimately, compared to traditional parallax image synthesis methods that are prone to inconsistent pixel ratios, this invention achieves uniform pixel size by displaying preprocessed parallax image pixels across cylindrical lenses.

Claims

1. An image processing method for pixel size uniformity, characterized in that: This image processing method for pixel size uniformity includes the following steps: (1) Image resolution preprocessing Preprocess all N original disparity images. Let the resolution of the original disparity images be X×Y (X rows, Y columns of pixels). Then, transform the resolution of each original disparity image into a preprocessed disparity image of resolution J×K, where J and K satisfy the following conditions: J=O; K = P / 2N + 1; Where O is the number of pixel rows in the parallax composite image, P is the number of pixel columns in the parallax composite image, and N is the number of viewpoints on the 3D display. (2) Pixel arrangement processing For the i-th column of the n-th preprocessed disparity image, the pixels are placed in the j-th and k-th columns of the disparity composite image, where j = (i-2)×2N+2n and k = (i-1)×2N+2n-1, respectively, and 1≤j≤P, 1≤k≤P and n∈{1,2,3,…,N}.

2. The image processing method for pixel size uniformity as described in claim 1, characterized in that: Set the edge column of the parallax composite image to 0.

Images