Machine Vision Image Compression via Redundancy-Aware Resampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image/video compression techniques focus on human perception, failing to optimize image data for machine vision tasks, leading to inefficient data transmission and storage for applications like autonomous driving and intelligent transportation.
Innovation Solution
Perform pre-analysis to identify temporal and spatial redundancies in image data, followed by down-sampling and encoding, and optionally up-sampling for machine vision tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional image/video compression techniques are used to ensure human perception quality, then image quality is improved, but compression efficiency for machine vision tasks deteriorates
Solution Approach 1:
The patent changes the compression parameters and approach from human-perception-based to machine-vision-based optimization. By analyzing temporal and spatial redundancy patterns specific to machine vision tasks, the system adjusts compression parameters to maximize machine vision performance while reducing data volume, thereby resolving the contradiction between traditional quality preservation and new efficiency requirements
Solution Approach 2:
The patent performs pre-analysis of image data to identify temporal and spatial redundancy patterns before compression. This preliminary action enables the system to optimize the compression process specifically for machine vision tasks, improving both compression efficiency and downstream task performance simultaneously
2Productivity
If image data is compressed without considering machine vision requirements, then transmission and storage efficiency is improved, but machine vision task performance deteriorates
Solution Approach 1:
The patent applies different compression strategies to different aspects of image data based on their importance for machine vision tasks. By analyzing local redundancy patterns and task requirements, the system optimizes compression parameters specifically for machine vision applications, ensuring both efficient data handling and maintained task performance
Solution Approach 2:
The patent incorporates feedback mechanisms that evaluate compression results against machine vision task performance. This feedback loop enables continuous optimization of compression parameters to ensure that compression efficiency improvements do not compromise downstream task accuracy
3Productivity
If pre-analysis based resampling compression is performed, then compression efficiency for machine vision is improved, but processing complexity increases
Solution Approach 1:
The patent segments the compression process into distinct stages: pre_analysis, resampling, and compression. By dividing the complex task into manageable segments, the system can optimize each stage independently while maintaining overall efficiency. The pre_analysis stage identifies redundancy patterns, the resampling stage applies appropriate transformations, and the compression stage finalizes the data reduction
Data Source
AI summary
An image data encoding method is provided. The image data encoding method includes receiving an input image and a quantization parameter (QP); performing pre-analysis of the input image based on the QP to obtain input image data for machine version; determining at least one of a temporal redundancy or a spatial redundancy of the input image data; determining, based on the at least one of temporal redundancy or spatial redundancy, whether to perform at least one of temporal down-sampling or spatial down-sampling; performing down-sampling of the input image data according to the determination, wherein the down-sampling includes at least one of temporal down-sampling or spatial down-sampling; and encoding the down-sampled image data based on the QP.