Image Block Processing with Extended Areas for Seamless Boundary Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image processing methods using deep learning networks face slow processing speeds due to hardware limitations, resulting in the need to divide images into blocks, which often lead to abnormal color spots and obvious boundaries when combining processed blocks.
Innovation Solution
An image processing method that divides images into blocks with effective and extended areas, processes each block using a target network model, extracts target effective areas, and combines them to generate a target image without filling redundant information, thereby eliminating abnormal color spots and boundaries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If an image is divided into multiple image blocks for processing, then processing speed is improved, but abnormal color spots and obvious boundaries appear at the boundaries of image blocks
Solution Approach 1:
The image is divided into multiple image blocks that are processed separately through the deep learning network. Each image block includes an effective image area and an extended area, allowing parallel processing while maintaining boundary quality through proper segmentation design
Solution Approach 2:
The image blocks are padded with extended areas before processing. This preliminary action ensures that when image blocks are combined, the boundaries align properly and redundant information filling is avoided, preventing abnormal color spots at boundaries
2Reliability
If redundant information is filled into image blocks for processing, then the deep learning network can process complete image data, but abnormal color spots appear at boundaries after combination
Solution Approach 1:
Different areas of the image block are treated differently: the effective image area contains the actual image data to be processed, while the extended area provides boundary context without requiring redundant information filling. This local differentiation ensures processing completeness while avoiding harmful color spots at boundaries
Solution Approach 2:
The patent extracts only the necessary effective image areas from the processed image blocks for final combination, excluding the extended areas that would contain redundant information. This extraction prevents abnormal color spots while maintaining processing reliability
Data Source
AI summary
An image processing method, an electronic device, and a non-transitory computer readable storage medium are provided. The method includes: dividing a to-be-processed image into blocks to obtain a plurality of image blocks. Each image block includes an effective image area and an extended area. The method further includes processing image data of each image block by using a target network model to obtain each target image block. The method also includes extracting target effective image areas in each target image block separately. The method additionally includes combining the target effective image areas to generate a target image.