Image Upscaling Neural Network Multiplexer Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image upscaling methods, such as bi-cubic and linear schemes, involve a large number of data operations and lack flexibility in adjusting the upscaling factor, leading to inefficient image resolution enhancement.
Innovation Solution
An apparatus comprising a cascade connection of convolutional neural network circuits and multiplexers, where each multiplexer integrates n*n feature images into a feature image with n times the resolution, allowing for flexible adjustment of the upscaling factor by arranging multiple multiplexers to achieve desired resolution enhancements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If standard upscaling schemes (bi-cubic, linear) are used, then image resolution is increased, but the number of data operations becomes large and the upscaling factor cannot be adjusted flexibly
Solution Approach 1:
The patent segments the upscaling process into multiple stages using a cascade structure with multiple convolutional neural network circuits and multiplexers. Each stage processes a portion of the upscaling task, allowing flexible combination to achieve different upscaling factors while reducing the computational burden on each individual stage.
Solution Approach 2:
The patent employs dynamic configuration where multiplexers can be selectively enabled or disabled based on the desired upscaling factor. This allows the system to adaptively adjust the number of operational stages to match the required upscaling factor, providing flexibility without permanently maintaining all possible configurations.
2Productivity
If standard upscaling schemes are used, then image resolution is increased, but computational efficiency is reduced due to large number of data operations
Solution Approach 1:
By dividing the upscaling computation into segmented stages with intermediate multiplexer operations, the patent reduces the computational complexity at each stage compared to traditional single-stage methods. This segmentation allows for more efficient parallel processing and reduces overall processing time.
Solution Approach 2:
The cascade structure performs preliminary processing at each stage, preparing intermediate results that are progressively refined. This preliminary action at multiple stages avoids the need for extensive single-stage computation, improving computational efficiency and reducing processing time.
Data Source
AI summary
The disclosure discloses an apparatus for upscaling an image, a method for training the same, and a method for upscaling an image, where a convolutional neural network circuit obtains feature images of the image, a multiplexer upscales the image by integrating every n*n feature images of an input signal into a feature image with a resolution which is n times the resolution of a feature image of the image, where n is an integer greater than 1.