Block-Based Image Compression With Neighborhood Context
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing deep learning-based image compression schemes encounter computational resource limitations and produce block artefacts when encoding high-resolution images by dividing them into smaller pixel blocks.
Innovation Solution
The proposed method uses a machine learning algorithm that incorporates block neighborhood data, including pixels outside the block's edges, to encode images, employing filtering techniques like Mean and Gaussian filters, and trains the algorithm to minimize a distortion metric using a rate-distortion cost function.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If images are divided into pixel blocks for encoding using deep learning models, then computational resource requirements are reduced, but visual quality deteriorates due to block effect artefacts
Solution Approach 1:
The image is divided into pixel blocks for encoding, which reduces computational resource requirements by processing smaller segments independently. This segmentation enables parallel processing and reduces the memory footprint during deep learning inference.
Solution Approach 2:
The patent combines the pixel block with its neighborhood pixels as input to the neural network. By merging surrounding context information with the target block, the model can generate smoother transitions at block boundaries while maintaining the computational efficiency of block-based processing.
2Object-affected harmful factors
If block neighborhood data is incorporated into the encoding process, then visual quality improves by reducing block effect, but computation complexity increases
Solution Approach 1:
The patent applies different processing strategies to different regions: blocks with neighborhood data for edge regions where block effects are most prominent, and standard block processing for interior regions. This local quality approach optimizes the balance between visual quality and computational complexity.
Solution Approach 2:
The patent incorporates partial neighborhood information (only certain surrounding pixels) rather than complete contextual data, achieving sufficient block effect mitigation without the full computational cost of processing all surrounding pixels.
Data Source
AI summary
A method for encoding image data of an image divided into a plurality of pixel blocks using a machine learning algorithm, is proposed, which comprises, by a computing platform comprising a processor configured for implementing the machine learning algorithm, for a block of the pixel blocks: obtaining a block neighborhood set of at least one pixel of the image, wherein the at least one pixel is located outside the block in a spatial neighborhood of a corresponding pixel located on an edge of the pixel block, and generating, by the machine learning algorithm configured for performing end-to-end image compression, a bitstream representing the encoded pixel block, by encoding the pixel block based on input data comprising the pixel block and the block neighborhood set.