Convolutional Neural Network Image Upscaling Hue Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods using deep neural networks, such as CNNs, fail to maintain hue consistency when scaling images, leading to hue distortion between input and output images.
Innovation Solution
An image processing method utilizing a convolutional neural network (CNN) that includes upscaling an input image, obtaining feature maps, generating gain and offset maps, and producing an output image by element-wise multiplication and addition, with the CNN trained to minimize hue differences between input and output images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If image scaling is performed using a deep convolutional neural network independently for each hue channel, then image resolution is improved, but hue distortion occurs between input and output images
Solution Approach 1:
The patent segments the image processing into two distinct stages: first, a CNN performs upscaling to improve resolution; second, a separate hue correction module processes the upscaled image to restore original hues. This segmentation allows each module to specialize in one function, preventing the hue distortion that occurs when a single network tries to handle both tasks simultaneously.
Solution Approach 2:
The patent introduces an intermediary hue correction module that acts as a mediator between the upscaled image and the final output. This intermediate processing step corrects the hue distortion introduced by the CNN upscaling, ensuring that the final image maintains both high resolution and accurate color representation.
2Productivity
If a single convolutional neural network performs both upscaling and hue correction, then processing efficiency is improved, but hue distortion increases
Solution Approach 1:
The patent divides the processing system into two specialized components: a CNN for efficient upscaling and a separate hue correction module for accurate color preservation. This segmentation maintains processing efficiency while improving hue accuracy, as each component can be optimized for its specific function without compromising the other.
Solution Approach 2:
The patent combines the upscaled image and the original image through a merging process where the hue correction module applies corrections based on the original image's hue information. This merging allows the system to leverage both the high resolution of the upscaled image and the accurate hues of the original image.
Data Source
AI summary
Provided is an image processing method including upscaling an input image to generate an upscaled image, obtaining a first feature map and a second feature map by inputting the upscaled image to a convolutional neural network and performing a convolution operation on the upscaled image with one or more kernels included in the convolutional neural network, obtaining a gain map by inputting the first feature map to a first convolutional layer, obtaining an offset map by inputting the second feature map to a second convolutional layer, and generating an output image, based on the upscaled image, the gain map, and the offset map, wherein the convolutional neural network is configured to be trained to reduce a difference between a hue of the input image and a hue of the output image.