Convolutional Neural Network Processor for Image Processing Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current convolutional neural networks process both luminance and chrominance information densely, leading to increased workload and reduced efficiency in image processing.
Innovation Solution
A convolutional neural network processor with a first convolutional unit extracting feature maps of different scales, a processing unit fusing noise information with the smallest feature map, and a second convolutional unit combining these feature maps to produce a processed image, utilizing sampling units for down-sampling and up-sampling, and interlayer connections for direct input between corresponding layers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a dense link architecture is adopted to process the image as a whole, then image processing can be performed, but the workload of image processing increases and processing efficiency reduces
Solution Approach 1:
The patent divides the image processing task into two separate processing paths: a first processing path that processes luminance information and a second processing path that processes chrominance information. This segmentation allows each path to handle only specific types of information, reducing the overall workload compared to processing the entire image densely, while still achieving complete image processing through the combination of both paths.
2Reliability
If both luminance and chrominance information are processed, then complete image processing is achieved, but the workload increases and efficiency reduces
Solution Approach 1:
The patent segments the processing of luminance and chrominance information into separate processing paths. The first processing path handles luminance information while the second processing path handles chrominance information. This ensures that both types of information are processed completely (maintaining reliability) while avoiding the inefficiency of processing them together in a dense architecture (improving productivity).
Solution Approach 2:
The patent applies different processing characteristics to different types of image information. The first processing path is optimized for luminance information with specific convolutional layer configurations, while the second processing path is optimized for chrominance information. This local optimization allows each path to process its specific type of information more efficiently, improving overall processing efficiency while maintaining complete processing of both luminance and chrominance information.
Data Source
AI summary
The present disclosure discloses a convolutional neural network processor, an image processing method and an electronic device. The method includes: receiving, by the first convolutional unit, the input image to be processed, extracting the N feature maps with different scales in the image to be processed, sending the N feature maps to the second convolutional unit, and sending the first feature map to the processing unit; fusing, by the processing unit, the received preset noise information and the first feature map, to obtain the second feature map, and sending the second feature map to the second convolutional unit; and fusing, by the second convolutional unit, the received N feature maps with the second feature map to obtain the processed image.