# US 2016/0260198 A1 · Pixel-based method and apparatus of obtaining downsampled image

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2016/0260198 A1
- Document ID: US20160260198A1
- Status: Published application
- Kind: A1
- Application number: 14/395,519
- Filed: 2014-09-09
- Published: 2016-09-08
- Earliest claimed: 2014-08-27
- Assignees: Shenzhen China Star Optoelectronics Technology Co., Ltd.
- Applicants: Shenzhen China Star Optoelectronics Technology Co., Ltd.
- Inventors: Houliang HU; Hao LI; Li-wei CHU
- Web page: https://patentgrep.com/patent/US20160260198A1

## CPC

- G06T 3/40: Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06F 18/22: Matching criteria, e.g. proximity measures
- G06T 3/4053: based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Pixel-based method and apparatus of obtaining a downsampled image are provided. The method includes: constructing a virtual image by an image super-resolution technology according to a resolution of the downsampled image; determining relational expressions between pixels of the virtual image and pixels of the downsampled image; calculating a MSE between the pixels of the virtual image and pixels of the original image based on the relational expressions; determining a coefficient matrix between the pixels of the downsampled image and the pixels of the original image under the condition of the MSE being at a minimum value; determining the pixels of the downsampled image according to the pixels of the original image and the coefficient matrix; and outputting the downsampled image. Accordingly, the present invention can obtain sharp image and avoid color aliasing.

## Claims

### Claim 1

A pixel-based method of obtaining a downsampled image, wherein a resolution of an original image is 2M*2N, and a resolution of the downsampled image is M*N; the method comprising steps of:
constructing a virtual image by an image super-resolution technology according to the resolution M*N of the downsampled image, wherein a resolution of the virtual image is 2M*2N;
determining relational expressions between pixels of the virtual image and pixels of the downsampled image;
calculating a mean squared error between the pixels of the virtual image and pixels of the original image based on the relational expressions between the pixels of the virtual image and the pixels of the downsampled image;
determining a coefficient matrix between the pixels of the downsampled image and the pixels of the original pixels under the condition of the mean squared error between the pixels of the virtual image and the pixels of the original image being at a minimum value;
determining the pixels of the downsampled image according to the pixels of the original image and the coefficient matrix between the pixels of the downsampled image and the pixels of the original image;
outputting the downsampled image as per the pixels of the downsampled image;
wherein the step of constructing a virtual image by an image super-resolution technology according to the resolution M*N of the downsampled image comprises: constructing the virtual image by an image super-resolution interpolation method according to the resolution M*N of the downsampled image;
the step of constructing the virtual image by an image super-resolution interpolation method according to the resolution M*N of the downsampled image comprises:
setting P1, P2, P3, P4 as pixels of the downsampled image and A, B, C as pixels of the virtual image constructed by the image super-resolution interpolation method, wherein positional relationships among P1, P2, P3, P4, A, B, C are that: A is located at an intersection of a diagonal line connecting P1 and P4 and another diagonal line connecting P2 and P3, B is located at a midpoint of a line connecting P1 and P2, and C is located at a midpoint of a line connecting P1 and P3;
determining interpolation relational expressions between R, G, B subpixel values of the pixels A, B, C of the virtual image and R, G, B subpixel values of the pixels P1, P2, P3, P4 of the downsampled image as that: A=¼(P1+P2+P3+P4), B=½(P1+P2), C=½(P1+P3);
determining interpolation relations between R, G, B subpixel values of the pixels of the virtual image and R, G, B subpixel values of the pixels of the downsampled image;
calculating all the pixels of the downsampled image to construct the virtual image as per the interpolation relations between R, G, B subpixel values of the pixels of the virtual image and R, G, B subpixel values of the pixels of the downsampled image.

### Claim 2 (depends on 1)

The method as claimed in claim 1, wherein the step of determining the pixels of the downsampled image according to the pixels of the original image and the coefficient matrix between the pixels of the downsampled image and the pixels of the original image comprises:
sorting the magnitudes of coefficients between each of the pixels of the downsampled image and all the pixels of the original image according to the coefficient matrix between the pixels of the downsampled image and the pixels of the original image;
determining actually used coefficients for each of the pixels of the downsampled image and pixels of the original image corresponding to the actually used coefficients according to a sum of relatively large coefficients after sorting fell in the range of no less than 0.95 and no more than 1.05;
determining the pixels of the downsampled image according to the actually used coefficients for each of the pixels of the downsampled image and the pixels of the original image corresponding to the actually used coefficients.

### Claim 3

A pixel-based method of obtaining a downsampled image, wherein a resolution of an original image is 2M*2N, and a resolution of the downsampled image is M*N; the method comprising steps of:
constructing a virtual image by an image super-resolution technology according to the resolution M*N of the downsampled image, wherein a resolution of the virtual image is 2M*2N;
determining relational expressions between pixels of the virtual image and pixels of the downsampled image;
calculating a mean squared error between the pixels of the virtual image and pixels of the original image based on the relational expressions between the pixels of the virtual image and the pixels of the downsampled image;
determining a coefficient matrix between the pixels of the downsampled image and the pixels of the original pixels under the condition of the mean squared error between the pixels of the virtual image and the pixels of the original image being at a minimum value;
determining the pixels of the downsampled image according to the pixels of the original image and the coefficient matrix between the pixels of the downsampled image and the pixels of the original image;
outputting the downsampled image as per the pixels of the downsampled image

### Claim 4 (depends on 3)

The method as claimed in claim 3, wherein the step of constructing a virtual image by an image super-resolution technology according to the resolution M*N of the downsampled image comprises: constructing the virtual image by an image super-resolution interpolation method according to the resolution M*N of the downsampled image.

### Claim 5 (depends on 4)

The method as claimed in claim 4, wherein the step of constructing the virtual image by an image super-resolution interpolation method according to the resolution M*N of the downsampled image comprises:
setting P1, P2, P3, P4 as pixels of the downsampled image and A, B, C as pixels of the virtual image constructed by the image super-resolution interpolation method, wherein positional relationships among P1, P2, P3, P4, A, B, C are that: A is located at an intersection of a diagonal line connecting P1 and P4 and another diagonal line connecting P2 and P3, B is located at a midpoint of a line connecting P1 and P2, and C is located at a midpoint of a line connecting P1 and P3;
determining interpolation relational expressions between R, G, B subpixel values of the pixels A, B, C of the virtual image and R, G, B subpixel values of the pixels P1, P2, P3, P4 of the downsampled image as that: A=¼(P1+P2+P3+P4), B=½(P1+P2), C=½(P1+P3);
determining interpolation relations between R, G, B subpixel values of the pixels of the virtual image and R, G, B subpixel values of the pixels of the downsampled image;
calculating all the pixels of the downsampled image to construct the virtual image as per the interpolation relations between R, G, B subpixel values of the pixels of the virtual image and R, G, B subpixel values of the pixels of the downsampled image.

### Claim 6 (depends on 3)

The method as claimed in claim 3, wherein the step of determining the pixels of the downsampled image according to the pixels of the original image and the coefficient matrix between the pixels of the downsampled image and the pixels of the original image comprises:
sorting the magnitudes of coefficients between each of the pixels of the downsampled image and all the pixels of the original image according to the coefficient matrix between the pixels of the downsampled image and the pixels of the original image;
determining actually used coefficients for each of the pixels of the downsampled image and pixels of the original image corresponding to the actually used coefficients according to a sum of relatively large coefficients after sorting fell in the range of no less than 0.95 and no more than 1.05;
determining the pixels of the downsampled image according to the actually used coefficients for each of the pixels of the downsampled image and the pixels of the original image corresponding to the actually used coefficients.

### Claim 7

A pixel-based apparatus of obtaining a downsampled image, wherein a resolution of an original image is 2M*2N, and a resolution of the downsampled image is M*N; the apparatus comprising:
a virtual image constructing module, configured to construct a virtual image by an image super-resolution technology according to the resolution M*N of the downsampled image, wherein a resolution of the virtual image is 2M*2N;
a relational expression determining module, configured to determine relational expressions between pixels of the virtual image and pixels of the downsampled image;
a mean squared error calculating module, configured to a mean squared error between the pixels of the virtual image and pixels of the original image based on the relational expressions between the pixels of the virtual image and the pixels of the downsampled image;
a coefficient matrix determining module, configured to determine a coefficient matrix between the pixels of the downsampled image and the pixels of the original image under the condition of the mean squared error between the pixels of the virtual image and the pixels of the original image being at a minimum value;
a pixel determining module, configured to determine the pixels of the downsampled image according to the pixels of the original image and the coefficient matrix between the pixels of the downsampled image and the pixels of the original image;
an image output module, configured to output the downsampled image as per the pixels of the downsampled image.

### Claim 8 (depends on 7)

The apparatus as claimed in claim 7, wherein the virtual image constructing module is configured to construct the virtual image by an image super-resolution interpolation method according to the resolution M*N of the downsampled image.

### Claim 9 (depends on 8)

The apparatus as claimed in claim 8, wherein the virtual image constructing module comprises:
a setting unit, configured to set P1, P2, P3, P4 as pixels of the downsampled image and A, B, C as pixels of the virtual image constructed by the image super-resolution interpolation method, wherein positional relationships among P1, P2, P3, P4, A, B, C are that A is located at an intersection of a diagonal line connecting P1 and P4 and another diagonal line connecting P2 and P3, B is located at a midpoint of a line connecting P1 and P2, and C is located at a midpoint of a line connecting P1 and P3;
an interpolation relational expression determining unit, configured to determine interpolation relational expressions between R, G, B subpixel values of the pixels A, B, C of the virtual image and R, G, B subpixel values of the pixels P1, P2, P3, P4 of the downsampled image as that A=¼(P1+P2+P3+P4), B=½(P1+P2), C=½(P1+P3);
an interpolation relations determining unit, configured to determine interpolation relations between R, G, B subpixel values of the pixels of the virtual image and R, G, B subpixel values of the pixels of the downsampled image;
a virtual image constructing unit, configured to perform calculation on all the pixels of the downsampled image to construct the virtual image as per the interpolation relations between R, G, B subpixel values of the pixels of the virtual image and R, G, B subpixel values of the pixels of the downsampled image.

### Claim 10 (depends on 7)

The apparatus as claimed in claim 7, wherein the pixel determining module comprises:
a sorting unit, configured to sort the magnitudes of coefficients between each of the pixels of the downsampled image and all the pixels of the original image according to the coefficient matrix between the pixels of the downsampled image and the pixels of the original image;
an actually used coefficient determining unit, configured to determine actually used coefficients for each of the pixels of the downsampled image and pixels of the original image corresponding to the actually used coefficients according to a sum of relatively large coefficients after sorting fell in the range of no less than 0.95 and no more than 1.05;
a pixel determining unit, configured to determine the pixels of the downsampled image according to the actually used coefficients for each of the pixels of the downsampled image and pixels of the original image corresponding to the actually used coefficients.
