CN112435171B

CN112435171B - Reconstruction method of image resolution

Info

Publication number: CN112435171B
Application number: CN202110116306.3A
Authority: CN
Inventors: 朱金伟
Original assignee: Hangzhou Xitong Intelligent Technology Co ltd
Current assignee: Hangzhou Xitong Intelligent Technology Co ltd
Priority date: 2021-01-28
Filing date: 2021-01-28
Publication date: 2021-04-20
Anticipated expiration: 2041-01-28
Also published as: CN112435171A

Abstract

The invention relates to the technical field of image processing, in particular to a method for reconstructing image resolution, which comprises the steps of projecting pixels of an original image into an enlarged image in an equal proportion, assigning blank pixels inserted into reference pixels according to the projected reference pixels, and enabling the assignment of the blank pixels to be closer to the values of the reference pixels closer to the blank pixels during assignment so as to enable the overall sharpness of the image to be higher and avoid the excessive blurring of the image after resolution reconstruction; meanwhile, the edge position is assigned in an exponential mode, so that the RGB value of the blank pixel at the edge position can be rapidly converged, the value of the blank pixel at the edge position can be closer to the value of the reference pixel at the edge, and edge blurring and sawtooth are avoided.

Description

Reconstruction method of image resolution

Technical Field

The invention relates to the technical field of image processing, in particular to a reconstruction method of image resolution.

Background

The resolution of an image refers to the resolution of an imaging system to the details of the image, and is one of the important indicators for measuring the quality of the image. The high-resolution image can provide abundant detailed information, and with the continuous progress of economy, science and technology and civilization, the demand of the high-resolution image is increasing in various fields such as medicine, safety, entertainment and the like. For example, a physician may wish to identify a lesion by high resolution CT or B-ultrasound imaging; the public security department hopes to identify the identity of a suspect or vehicle information through a high-resolution monitoring image; entertainment businesses desire to achieve more realistic and fine visual effects for viewers through high-resolution video.

The most direct method for improving the resolution is to increase the hardware resolution of the digital image acquisition system, and the method is mainly considered from the aspects of improving the resolution of an image sensor and the resolution of a lens. However, the hardware method has technical bottleneck, and is expensive and difficult to popularize and apply. The method of improving the resolution of an image by a software method is called a super-resolution reconstruction technique, which reconstructs one or more high-resolution images using complementary information that may exist between low-resolution images. The super-resolution reconstruction technology is slowly developed at the beginning, research is mainly focused on the trial and exploration of using some classical methods (such as interpolation, regularization, least square method and the like) for the technology, and the reconstruction effect is often not ideal. Recently, with the proposition and development of theories and methods such as graph cut algorithm, sparse representation, deep learning and the like, the super-resolution reconstruction technology makes great progress under the influence of the new theories and new methods, and the combination of the new theories and the super-resolution reconstruction technology remarkably improves the reconstruction effect and the reconstruction speed.

Analyzing the existing reconstruction methods, some interpolation methods have poor effect when improving the resolution, and blur and sawtooth often occur.

Disclosure of Invention

In view of the above, the present invention provides a method for reconstructing image resolution, which can solve the problems in the background art.

The invention relates to a reconstruction method of image resolution, which comprises the following steps:

acquiring an RGB value of each pixel of an image through a computer, establishing an x-y coordinate system by taking the lower left corner of the image as a zero point, and corresponding the coordinate of each pixel point to the RGB value to obtain a pixel reference value A (x, y) of the original image, wherein the A (x, y) is the RGB value of the pixel of the original image;

establishing an x ' -y ' coordinate system of a reconstructed image by taking the lower left corner of the blank image as a zero point, projecting a pixel reference value of an original image into the coordinate system of the reconstructed image according to a resolution reconstruction proportion n, and obtaining a pixel value A ' (nx, ny) of the reconstructed image, wherein A ' (x, y) ═ A ' (nx, ny) is an RGB value of a pixel of the reconstructed image;

the blank pixels B (x ', y') between the projected pixels of the coordinate system of the reconstructed image are assigned according to the following assignment rule:

(1) when the blank pixel B (x ', y') is located on a straight line connecting the pixels a1'(x1', y1') and a2' (x2', y2') of any two reconstructed images, the blank pixel B (x ', y') is located on a straight line connecting the pixels a1'(x1', y1') and a2' (x2', y2') of any two reconstructed images

K is an offset;

(2) when the blank pixel B (x ', y') is not on the pen-direct line of any two pixels of the reconstructed image, the blank pixel B (x ', y') is not on the pen-direct line of any two pixels of the reconstructed image

Here, the pixel A3(x3', y3') is a pixel of a reconstructed image closest to the blank pixel B (x ', y'), and m is a pixel distance between the pixel A3(x3', y3') and the blank pixel B (x ', y').

Further, the offset K is used to adjust the pixel bias value of the reconstructed image.

Further, the pixel RGB values of the reconstructed image are converted into gray values and then assigned, and the formula of gray value conversion is gray ═ R × 0.3+ G × 0.59+ B × 0.11.

Further, any two adjacent blank pixels B1(x1', y1') and B2(x2', y2') assigned are satisfying: when B1(x1', y1') -B2(x2', y2') ≧ α, the assigned blank pixel B1(x1', y1') is decreased or the assigned blank pixel B2(x2', y2') is increased in value so that B1(x1', y1') -B2(x2', y2') < α, α is a preset threshold.

Further, n²The resolution reconstruction multiplying power is n less than or equal to 4, wherein n is a natural number.

The invention has the beneficial effects that: according to the image resolution reconstruction method, the pixels of the original image are projected into the amplified image in an equal proportion, the blank pixels inserted into the reference pixels are assigned according to the projected reference pixels, and when the blank pixels are assigned, the assignment of the blank pixels can be closer to the values of the reference pixels closer to the blank pixels, so that the overall sharpness of the image is higher, and the image after resolution reconstruction is prevented from being excessively blurred; meanwhile, the edge position is assigned in an exponential mode, so that the RGB value of the blank pixel at the edge position can be rapidly converged, the value of the blank pixel at the edge position can be closer to the value of the reference pixel at the edge, and edge blurring and sawtooth are avoided.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and it will be apparent to those skilled in the art that other related drawings can be obtained from the drawings without inventive effort.

FIG. 1 is a flow chart of a method of the present invention;

FIG. 2 is a coordinate system reconstruction for pixel processing of the present invention;

FIG. 3 is a schematic representation of the reconstruction assignment of the pixel process of the present invention;

FIG. 4 is a schematic diagram of the present invention before and after resolution reconstruction.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations.

Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

As shown in fig. 1: the method for reconstructing the image resolution of the embodiment includes the steps:

establishing an x ' -y ' coordinate system of a reconstructed image by taking the lower left corner of the blank image as a zero point, projecting a pixel reference value of an original image into the coordinate system of the reconstructed image according to a resolution reconstruction proportion n to obtain a pixel value A ' (nx, ny) of the reconstructed image, wherein A ' (nx, ny) is an RGB value of a pixel of the reconstructed image, and A (x, y) ═ A ' (nx, ny), n²For the resolution reconstruction multiplying factor, n is less than or equal to 4, that is, the reconstructed image pixels can only be 16 times of the original image at most, if the multiplying factor is too large, the number of blank pixels inserted between the reference pixels is too large, and the image can be seriously distorted after being assigned through operation.

Assigning blank pixels B (x ', y ') between projected pixels of the coordinate system of the reconstructed image, before assigning, for the convenience of calculation, converting A (x, y) and A ' (nx, ny) into gray values, before the gray values are 0-255, and converting the gray values into gray values of R0.3 + G0.59 + B0.11 according to the following assignment rule:

K is an offset amount used to adjust the pixel polarization value of the reconstructed image, the assignment of the blank pixel is determined according to the distance from the blank pixel B (x ', y') closer to the reference pixel a1'(x1', y1'), the closer to the reference pixel a1' (x1', y1'), the closer to the reference pixel a2'(x2', y2'), the closer to the reference pixel a2' (x2', y2') the blank pixel B (x ', y') is assigned, to the reference pixels a1'(x1', y1') and a2' (x2', y 2').

Wherein the pixel A3(x3', y3') is the pixel of the reconstructed image closest to the blank pixel B (x ', y'), and m is the pixel distance between the pixel A3(x3', y3') and the blank pixel B (x ', y'), and the assignment of the blank pixel B (x ', y') is closer to the reference pixel A3(x3', y3') and exponentially changed as the blank pixel B (x ', y') is closer to the reference pixel A3(x3', y3'), so that the edge of the image has a narrow gradient color band.

In addition, in order to avoid the phenomenon of stripes, lines and the like caused by the large color difference between blank pixels after the blank pixels are assigned after the resolution reconstruction occurs, in this embodiment, if any two adjacent assigned blank pixels B1(x1', y1') and B2(x2', y2') satisfy: when B1(x1', y1') -B2(x2', y2') ≧ α, the assigned blank pixel B1(x1', y1') is decreased or the assigned blank pixel B2(x2', y2') is increased in value so that B1(x1', y1') -B2(x2', y2') < α, specifically, an average value algorithm or a threshold value algorithm may be used, for example, an interpolated value β is defined upon occurrence

B1(x1', y1') -B2(x2', y 2'). gtoreq.alpha,

B1(x1'，y1')'＝B1(x1'，y1')-β；

B2'(x2'，y2')＝B2(x2'，y2')+β。

so that the difference is less than the threshold a.

The threshold β needs to be set according to the pixel RGB value distribution of a specific image, and is generally 5 to 15.

And finally, restoring the gray value into an RGB value.

Specifically, as shown in fig. 2-3, in the present embodiment, n is 3, according to the reconstruction method in the present invention, a coordinate system x-y is first established using the lower left corner of the original image as a zero point, for convenience, the present embodiment takes 4 pixels at the middle position for explanation, the coordinates of the 4 pixels are a1 (10000), a2(10001,10000), A3(10001,10000), and a4 (10001), and the gray values of the four pixels after gray value conversion are: a1 (10000) ═ 120; a2(10001,10000) ═ 210; a3(10001,10000) ═ 60; a4 (10001) ═ 150, for ease of illustration, the bias values for the four adjacent pixels are larger;

(1) taking a blank image with pixel initial values of 255 (namely white), establishing a coordinate system x '-y' of a reconstructed image by taking a pixel point at the lower left corner as a zero point, wherein coordinates of the four projected pixels are respectively A1 '(30000), A2' (30003,30000), A3'(30003,30000) and A4' (30003), and at the moment, 12 white pixel points B1-B12 are generated among the four reference pixels, wherein the 12 white pixel points are all located on straight connecting lines of the four reference pixels, and in addition, some white pixels are located at the edges of the four reference pixels, and are located on connecting lines of the four reference pixels and other reference pixels, which is not to be said at first;

where B1 and B2 are between A1 'and A2', with coordinates B1(30001,30000), B2(30002,30000), respectively, brought into the assignment formula, the offset value K is first set to 0,

the gray values of B1 and B2 are 150 and 180;

similarly, the gray-scale values corresponding to B3, B6, B7, B10, B11 and B12 are determined as 100, 190, 80, 170, 70 and 120, and the gray-scale values are substituted into B3, B6, B7, B10, B11 and B12;

wherein, B5 and B8 are located between A2 'and A3', B4 and B9 are located between A1 'and A4', and are substituted into the assignment formula, so that the gray values of B4, B5, B8 and B9 are respectively: 130. 160, 110 and 140, and bringing the gray values into B4, B5, B8 and B9;

(2) when the white pixels B13-B32 of the edge portions of the four reference pixels are the most marginal pixels, the edge assignment formula is used

Assigning values to B13-B32, and substituting gray values of A1', B1, B2, A2', B3, B7 and A3' into a formula, wherein B15 and B21, B16 and B22, B17 and B23, B18 and B24 are respectively nearest to A1', B1, B2 and A2', and the gray values of B21-B24 are respectively: 44.1, 55.2, 66.2 and 77.3, the grey values of B15-B18 are 16.2, 20.3, 24.4 and 28.4, respectively;

b13, B14, B19 and B20 are nearest to A1' and the gray values are 16.2, 16.2 and 44.1 respectively;

the gray values of B25-B32 are 16.2, 44.1, 13.5, 36.8, 10.8, 29.4, 8.1 and 22.1 respectively;

at the moment, an obvious gradual change gray black edge appears at the edge of the image where the four reference pixels are located, the whole image is surrounded, the gradual change area is narrow, edge blurring is avoided, the gradual change black edge can be changed into a gradual change white edge if needed, and only the edge needs to be changedThe assignment formula becomes B (x ', y') -a 3(x3', y3') × e^mWhen B (x ', y') is greater than 255, 255 is required.

It should be noted that, in the actual image, it rarely occurs that the gray values of the adjacent pixels are greatly different, so that the gray values of the filling pixels are not always greatly different as shown in the embodiment, and the excessive effect of the actual filling pixels is more natural.

And finally, if the reconstructed picture has a color cast phenomenon, assigning a value to the offset value K to enable the reconstructed picture to be more natural, and removing the gray-black gradient edge at the edge of the reconstructed picture by cutting, wherein the final effect is as shown in FIG. 4 (the left side is before reconstruction, and the right side is after reconstruction).

According to the image resolution reconstruction method, the pixels of the original image are projected into the amplified image in an equal proportion, the blank pixels inserted into the reference pixels are assigned according to the projected reference pixels, and when the blank pixels are assigned, the assignment of the blank pixels can be closer to the values of the reference pixels closer to the blank pixels, so that the overall sharpness of the image is higher, and the image after resolution reconstruction is prevented from being excessively blurred; meanwhile, the edge position is assigned in an exponential mode, so that the RGB value of the blank pixel at the edge position can be rapidly converged, the value of the blank pixel at the edge position can be closer to the value of the reference pixel at the edge, and edge blurring and sawtooth are avoided.

Finally, the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting, although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions may be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all of them should be covered in the claims of the present invention.

Claims

1. A method of reconstructing image resolution, characterized by: the method comprises the following steps:

when the blank pixel B (x ', y') is located on a straight line connecting the pixels a1'(x1', y1') and a2' (x2', y2') of any two reconstructed images, the blank pixel B (x ', y') is located on a straight line connecting the pixels a1'(x1', y1') and a2' (x2', y2') of any two reconstructed images

K is an offset;

when the blank pixel B (x ', y') is not on the pen-direct line of any two pixels of the reconstructed image, the blank pixel B (x ', y') is not on the pen-direct line of any two pixels of the reconstructed image

2. A method of image resolution reconstruction as claimed in claim 1, wherein: the offset K is used to adjust the pixel bias value of the reconstructed image.

3. A method of image resolution reconstruction as claimed in claim 1, wherein: and converting the pixel RGB value of the reconstructed image into a gray value, and then assigning the gray value, wherein the gray value is converted into a formula of R0.3 + G0.59 + B0.11.

4. A method of image resolution reconstruction as claimed in claim 1, wherein: any two adjacent blank pixels B1(x1', y1') and B2(x2', y2') assigned are satisfied: b1(x1', y1') -B2(x2', y2') ≧ α, the assigned blank pixel B1(x1', y1') is decreased or the assigned blank pixel B2(x2', y2') is increased in value so that B1(x1', y1') -B2(x2', y2') < α, α preset threshold.

5. A method of image resolution reconstruction as claimed in claim 1, wherein: n is²The resolution reconstruction multiplying power is n less than or equal to 4, and n is a natural number.