Method for processing image, electronic device and storage medium

Patent metadata

Field	Value
Publication number	US-12482069-B2
Application number	US-202217888824-A
Country	US
Kind code	B2
Filing date	Aug 16, 2022
Priority date	Sep 30, 2021
Publication date	Nov 25, 2025
Grant date	Nov 25, 2025

Abstract

A method for processing an image, an electronic device, and a storage medium are provided. The method includes: acquiring an input image, the input image including a plurality of channels, and each channel of the plurality of channels including a plurality of pixel points; performing pixel extractions in parallel using a plurality of dedicated processing units on each channel of the input image, to obtain pixel point data for each corresponding pixel point of each channel; and splicing the pixel point data for each corresponding pixel point of each channel to obtain an output image.

First claim

The invention claimed is:

1 . A method for processing an image, the method comprising:
acquiring an input image, the input image comprising a plurality of channels, and each channel of the plurality of channels comprising a plurality of pixel points, wherein the plurality of channels comprises a first channel comprising a plurality of rows of pixel points; performing pixel extractions in parallel using a plurality of dedicated processing units on each channel of the input image, to obtain pixel point data for each corresponding pixel point of each channel; and splicing the pixel point data for each corresponding pixel point of each channel to obtain an output image, wherein splicing the pixel point data for each corresponding pixel point of each channel comprises: generating a given matrix comprising a plurality of rows and columns, wherein the generating the given matrix comprises:
generating a single row of pixel points in the given matrix, by arranging pixel points of a second row of the first channel immediately following pixel points of a first row of the first channel, wherein pieces of the pixel point data for the plurality of channels are arranged in respective rows or columns of the given matrix, and the single row of the given matrix comprises data of all pixel points of the first row and the second row of the first channel of the input image.

2 . The method according to claim 1 , wherein performing the pixel extractions in parallel comprises:
adding paddings to an edge of at least one side of the plurality of channels of the input image to obtain a plurality of padded channels; and performing the pixel extractions in parallel using the plurality of dedicated processing units on each of the plurality of padded channels, to obtain the pixel point data for each corresponding pixel point of each padded channel.

3 . The method according to claim 2 , further comprising:
determining a number of paddings to be added based on a size of the input image and a width of a convolution kernel for performing the pixel extraction.

4 . The method according to claim 1 , wherein performing the pixel extractions in parallel comprises:
determining, based on a number of channels comprised in the input image, a number of dedicated processing units for performing the pixel extractions in parallel.

5 . The method according to claim 1 , wherein performing the pixel extractions in parallel comprises:
determining a pixel to be extracted during the pixel extraction by each dedicated processing unit of the plurality of dedicated processing units based on at least one of: a horizontal stride of a convolution kernel for performing the pixel extraction; and a vertical stride of the convolution kernel for performing the pixel extraction.

6 . The method according to claim 1 , wherein the plurality of channels comprises a second channel comprising a plurality of rows, a second row of the plurality of rows of the given matrix comprises data of all pixel points of the plurality of rows of the second channel, and the method further comprises: multiplying each row of the given matrix with a corresponding block of a convolution kernel comprising a plurality of rows.

7 . An electronic device, comprising:
at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions, when executed by the at least one processor, cause the at least one processor to perform operations comprising: acquiring an input image, the input image comprising a plurality of channels, and each channel of the plurality of channels comprising a plurality of pixel points, wherein the plurality of channels comprises a first channel comprising a plurality of rows of pixel points; performing pixel extractions in parallel using a plurality of dedicated processing units on each channel of the input image, to obtain pixel point data for each corresponding pixel point of each channel; and splicing the pixel point data for each corresponding pixel point of each channel to obtain an output image, wherein splicing the pixel point data for each corresponding pixel point of each channel comprises: generating a given matrix comprising a plurality of rows and columns, wherein the generating the given matrix comprises:
generating a single row of pixel points in the given matrix, by arranging pixel points of a second row of the first channel immediately following pixel points of a first row of the first channel, wherein pieces of the pixel point data for the plurality of channels are arranged in respective rows or columns of the given matrix, and the single row of the given matrix comprises data of all pixel points of the first row and the second row of the first channel of the input image.

8 . The electronic device according to claim 7 , wherein performing the pixel extractions in parallel comprises:
adding paddings to an edge of at least one side of the plurality of channels of the input image to obtain a plurality of padded channels; and performing the pixel extractions in parallel using the plurality of dedicated processing units on each of the plurality of padded channels, to obtain the pixel point data for each corresponding pixel point of each padded channel.

9 . The electronic device according to claim 8 , wherein the operations further comprise:
determining a number of paddings to be added based on a size of the input image and a width of a convolution kernel for performing the pixel extraction.

10 . The electronic device according to claim 7 , wherein performing the pixel extractions in parallel comprises:
determining, based on a number of channels comprised in the input image, a number of dedicated processing units for performing the pixel extractions in parallel.

11 . The electronic device according to claim 7 , wherein performing the pixel extractions in parallel comprises:
determining a pixel to be extracted during the pixel extraction by each dedicated processing unit of the plurality of dedicated processing units based on at least one of: a horizontal stride of a convolution kernel for performing the pixel extraction; and a vertical stride of the convolution kernel for performing the pixel extraction.

12 . A non-transitory computer readable storage medium storing computer instructions, wherein the computer instructions when executed by a computer cause the computer to perform operations comprising:
acquiring an input image, the input image comprising a plurality of channels, and each channel of the plurality of channels comprising a plurality of pixel points, wherein the plurality of channels comprises a first channel comprising a plurality of rows of pixel points; performing pixel extractions in parallel using a plurality of dedicated processing units on each channel of the input image, to obtain pixel point data for each corresponding pixel point of each channel; and splicing the pixel point data for each corresponding pixel point of each channel to obtain an output image, wherein splicing the pixel point data for each corresponding pixel point of each channel comprises: generating a given matrix comprising a plurality of rows and columns, wherein the generating the given matrix comprises:
generating a single row of pixel points in the given matrix, by arranging pixel points of a second row of the first channel immediately following pixel points of a first row of the first channel, wherein pieces of the pixel point data for the plurality of channels are arranged in respective rows or columns of the given matrix, and the single row of the given ma

Assignees

• Kunlunxin Tech Beijing Company Limited

Inventors

• Liang Zihao
• Ouyang Jian
• Qi Wei
• Wang Jing

Classifications

• G06T2207/20221

  Image fusion; Image merging · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06T2200/32

  involving image mosaicing · CPC title

• G06T2207/20084

  Artificial neural networks [ANN] · CPC title

• G06T2207/20081

  Training; Learning · CPC title