Selective Patch Fusion in Image Processing Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing visual multilayer perceptron model requires a large computational workload due to the need to fuse all image patches, leading to inefficient and prolonged image processing.
Innovation Solution
An image processing method that evaluates patches based on their importance and selectively fuses only the most important patches, reducing the computational workload by omitting less important patches.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all patches are fused in the visual multilayer perceptron model, then complete image content is processed, but computational workload becomes excessively large
Solution Approach 1:
The patent extracts and removes less important patches from the set of all patches before fusion operations. By evaluating patch importance and selectively excluding low-importance patches, the model reduces computational workload while preserving essential image content, directly resolving the contradiction between processing completeness and efficiency
Solution Approach 2:
The patent applies different treatment to different patches based on their local importance characteristics. Important patches undergo full fusion processing while less important patches are excluded, creating a non-uniform processing strategy that optimizes the balance between completeness and efficiency by adapting to local patch qualities
2Reliability
If all patches are fused in the visual multilayer perceptron model, then comprehensive image features are obtained, but total processing duration becomes excessive
Solution Approach 1:
The patent extracts and removes less important patches from the fusion process based on importance evaluation. This extraction of non-essential patches reduces the number of operations required, thereby shortening total processing duration while maintaining accurate feature extraction from the remaining important patches
Solution Approach 2:
The patent applies partial fusion action by fusing only the most important patches rather than all patches. This partial action approach achieves sufficient feature extraction accuracy for visual tasks while significantly reducing processing time, as the essential features are captured without the overhead of processing all patches
3Reliability
If all patches are fused in the visual multilayer perceptron model, then complete image information is processed, but computational efficiency deteriorates
Solution Approach 1:
The patent extracts and eliminates less important patches from the computational process before fusion. This extraction reduces the number of computational operations required while preserving the accuracy of image processing by maintaining all important patches, thus improving computational efficiency without sacrificing reliability
Solution Approach 2:
The patent implements differentiated processing where important patches receive full fusion processing and less important patches are excluded. This local quality approach ensures that computational resources are focused on patches that contribute most to processing accuracy, thereby improving overall computational efficiency while maintaining processing reliability
Data Source
AI summary
This application discloses an image processing method and a related device thereof, to effectively reduce a computational workload of image processing, thereby shortening total duration of image processing, and improving image processing efficiency. The method in this application includes: after receiving N patches of a target image, a target model may first evaluate the N patches, to obtain evaluation values of the N patches. Next, the target model may select M patches from the N patches by using the evaluation values of the N patches as a selection criterion. Then, the target model may fuse the M patches, to obtain a fusion result of the M patches. Finally, the target model may perform a series of processing on the fusion result of the M patches, to obtain a processing result of the target image.