Neural Image Upscaling With Pixel Shuffle Preprocessing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image upscaling methods using neural networks face challenges in maintaining high-resolution performance across varying input resolutions and require significant computational resources, leading to inefficiencies in hardware complexity.
Innovation Solution
An electronic device employs a preprocessing method involving pixel shuffle and subtraction operations to optimize channel groups before inputting them into a neural network model, enhancing upscaling performance by minimizing data complexity and computational demands.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional neural network-based image upscaling methods are used, then high-resolution output can be achieved, but computational resources and hardware complexity increase significantly
Solution Approach 1:
The patent applies preliminary action by performing pixel shuffling and channel group reorganization before the image upscaling process. The method rearranges pixels into channel groups and subtracts reference values to create optimized input data that reduces computational complexity during the neural network processing stage, while maintaining high-resolution output quality
Solution Approach 2:
The patent changes data parameters by transforming the input image data through pixel shuffling operations and channel group reorganization. By rearranging pixel arrangements and adjusting channel group configurations, the method creates optimized data representations that reduce the computational burden on hardware while preserving image quality information
2Device complexity
If pixel shuffle and subtraction operations are applied to optimize channel groups, then computational requirements are reduced, but processing steps increase
Solution Approach 1:
The patent performs pixel shuffling and channel group optimization as preliminary actions before the main neural network processing. By preparing and organizing data in advance through these preprocessing steps, the method reduces the computational complexity of subsequent processing operations, leading to more efficient overall processing despite the additional preliminary steps
Data Source
AI summary
Provided is an electronic device and method of operating same, the electronic device including: memory storing instructions; and a processor configured to execute the instructions to: obtain an upscaled image by upscaling a plurality of pixels of an image; obtain channel groups by unshuffling pixels of the upscaled image; update the channel groups by subtracting a value based on a position of a pixel included in each of the channel groups in the upscaled image from a pixel value included in each of the channel groups; generate high-resolution channel groups by inputting the updated channel groups to a neural network; obtain a processed image having a resolution that is the same as a resolution of the upscaled image by shuffling the high-resolution channel groups; and obtain a final image in which the image is upscaled by performing convolution on the processed image with a preset filter.