Spatial Frequency Transform Image Modification Neural Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current neural network-based image and video coding methods lack efficiency in processing multiple image channels simultaneously, leading to suboptimal processing times and increased memory demands due to the need for all channels to be processed by the same network, which restricts the overall performance and flexibility in image enhancement tasks.
Innovation Solution
The method involves processing a primary image channel and one or more secondary channels using separate spatial frequency transforms and neural networks, allowing for independent optimization of each network's coefficients without affecting the other, and enabling different kernel usage for convolutional neural networks, thereby reducing processing time and memory requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all image channels are processed by the same neural network, then the network can capture inter-channel correlations, but the processing time and memory requirements increase significantly
Solution Approach 1:
The patent divides the processing of multiple image channels into separate neural network branches, where each branch processes one channel independently. This segmentation allows each network to be smaller and faster while the overall system maintains the ability to capture inter-channel correlations through the architecture design, resolving the contradiction between reliability and productivity.
2Reliability
If all image channels are processed by the same neural network, then comprehensive processing is achieved, but memory requirements increase
Solution Approach 1:
The patent segments the processing task across multiple independent neural networks, each handling a single image channel. This reduces the memory footprint of each individual network while maintaining comprehensive processing capability through the collective action of all networks, effectively resolving the memory requirements contradiction.
3Productivity
If separate neural networks are used for each image channel, then processing time is reduced and memory requirements decrease, but the ability to capture inter-channel correlations is compromised
Solution Approach 1:
The patent merges the output features from multiple separate neural networks that process different image channels into a unified representation. This combining mechanism allows the system to achieve both processing speed benefits from separate networks and inter-channel correlation capture through the merging operation, resolving the contradiction between productivity and reliability.
Data Source
AI summary
The present disclosure relates to image modification such as an image enhancement wherein the processing is at least partially based on neural networks. In particular, the image modification includes a multi-channel processing in which a primary channel is processed separately and secondary channels are processed based on the processed primary channel. The primary channel is processed based on a first spatial frequency transform to obtain a transformed primary channel and the secondary channel is processed based on a second spatial frequency transform to obtain a transformed secondary channel. The transformed primary channel is processed by means of a first neural network to obtain a modified transformed primary channel and the transformed secondary channel is processed based on the transformed primary channel by means of a second neural network to obtain a modified transformed secondary channel.