Stereo Image Compression via Disparity-Warped Feature Propagation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image compression methods for stereo camera pairs are inefficient as they compress images independently, leading to suboptimal use of storage space in applications like autonomous vehicles and three-dimensional movies.
Innovation Solution
A machine-learned image compression model that uses parametric skip functions and a conditional entropy model to jointly compress images from stereo pairs by propagating disparity-warped features between images, reducing the overall bitrate through enhanced encoding and decoding processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If images from stereo camera pairs are compressed separately using conventional methods, then the encoding process is simple and independent, but the storage efficiency is suboptimal and bitrate is higher
Solution Approach 1:
The patent merges the compression processes of two stereo images into a single joint compression framework. The encoder processes both images simultaneously, sharing computational resources and exploiting inter-image correlations to achieve better storage efficiency without proportionally increasing encoding complexity
Solution Approach 2:
The patent introduces a disparity dimension to the traditional single-image compression approach. By warping one image according to disparity maps and processing it alongside the other image, the system exploits the third dimension of depth information to reduce redundant data and improve compression ratios
2Quantity of substance
If joint compression of stereo images is implemented, then storage efficiency improves and bitrate reduces, but the encoding complexity increases
Solution Approach 1:
The patent performs preliminary disparity estimation and image warping before the main compression process. By pre-aligning the stereo images through disparity-based warping, the encoder reduces the complexity of subsequent joint compression operations and enables more efficient exploitation of inter-image correlations
Solution Approach 2:
The patent introduces disparity maps as intermediary data structures that mediate between the two stereo images. These disparity maps serve as a bridge to align and correlate features from both images, enabling efficient joint compression while managing computational complexity through structured intermediate representations
Data Source
AI summary
A machine-learned image compression model includes a first encoder configured to generate a first image code based at least in part on first image data. The first encoder includes a first series of convolutional layers configured to generate a first series of respective feature maps based at least in part on the first image. A second encoder is configured to generate a second image code based at least in part on second image data and includes a second series of convolutional layers configured to generate a second series of respective feature maps based at least in part on the second image and disparity-warped feature data. Respective parametric skip functions associated convolutional layers of the second series are configured to generate disparity-warped feature data based at least in part on disparity associated with the first series of respective feature maps and the second series of respective feature maps.