Compressed Image Noise Extraction Using Convolution Filters and DNN
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression technologies suffer from compressed artifacts, which degrade the quality of compressed images, and there is a need to effectively extract and reduce these artifacts to improve image quality.
Innovation Solution
The method involves using a convolution filter for down-convolution and up-convolution operations to separate structural and high-frequency noise regions in compressed images, followed by a deep neural network to remove compressed artifacts, thereby reconstructing the image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If image compression is applied to reduce data size, then storage efficiency and transmission speed are improved, but compressed artifacts occur that degrade image quality
Solution Approach 1:
The patent segments the compressed image into multiple frequency components using wavelet transform, separating low-frequency coefficients (containing structural information) from high-frequency coefficients (containing noise and artifacts). This segmentation allows selective processing of different frequency components to remove artifacts while preserving important image structures.
Solution Approach 2:
The patent extracts and removes high-frequency coefficients that correspond to noise and compressed artifacts. By taking out these problematic high-frequency components and replacing them with interpolated values or zero, the method eliminates artifacts while maintaining the compressed image's compact representation.
2Object-affected harmful factors
If conventional denoising methods are applied to reduce compressed artifacts, then image quality improves slightly, but computational complexity increases significantly
Solution Approach 1:
The patent performs wavelet transform and frequency component separation as preliminary actions before applying denoising. By pre-processing the image to separate frequency components, the subsequent denoising operation only needs to process the high-frequency coefficients, significantly reducing the computational complexity compared to applying denoising to the entire image.
Solution Approach 2:
The patent applies different processing strategies to different frequency components. Low-frequency coefficients are processed differently from high-frequency coefficients, allowing optimized denoising for each component type. This local quality approach avoids uniform processing of the entire image, reducing overall computational complexity.
Data Source
AI summary
Provided is a method of reducing noise of a compressed image, the method including obtaining a first image by applying a convolution filter, which sequentially performs down-convolution and up-convolution corresponding to the down-convolution, to a compressed image of an original image, obtaining a second image by subtracting the first image from the compressed image, obtaining noise comprising high-frequency information and a compressed artifact of the compressed image from the second image, obtaining a third image by removing the compressed artifact by applying a deep neural network (DNN) for removing the compressed artifact to the noise, and reconstructing the compressed image by summing the first image and the third image.