Stereoscopic Image Compression via Disparity Value Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for compressing stereoscopic image pairs for 3D video encoding face inefficiencies, either resulting in significant inaccuracy with global disparity vectors or excessive overhead when using tile-based approaches.
Innovation Solution
A method that generates and reduces disparity values for stereoscopic image pairs by determining the most frequent disparity values and using them to compress data, allowing for efficient regeneration of the second image from the first, minimizing data transmission volume while maintaining accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If a single global disparity vector is used for the entire image, then data transmission volume is reduced, but accuracy of 3D image reconstruction deteriorates
Solution Approach 1:
The image is divided into multiple tiles, and each tile is assigned its own disparity vector(s). This segmentation allows local accuracy to be maintained while keeping the overall data volume manageable, as each tile only needs to store disparity information for its specific region rather than the entire image.
Solution Approach 2:
The patent changes the parameter of disparity representation from a single global value to multiple local values organized in a hierarchical structure. By allowing different numbers of disparity vectors per tile and using variable precision representations, the system adapts the disparity data structure to match the actual complexity of each image region.
2Measurement precision
If tiles are divided and different disparity vectors are associated with each tile, then accuracy of 3D image reconstruction is improved, but data transmission volume and processing overhead increase excessively
Solution Approach 1:
Adjacent tiles that share the same disparity vector are merged into a single tile group, reducing the total number of disparity vectors needed. This merging eliminates redundancy between neighboring tiles while maintaining accurate disparity representation within each merged region.
Solution Approach 2:
The patent applies disparity vectors selectively - not every tile requires a unique disparity vector. By determining the minimum number of disparity vectors needed per tile based on actual image content complexity, the system avoids the excessive overhead of assigning vectors to all tiles uniformly.
3Manufacturing precision
If more disparity values are used to improve accuracy, then fidelity of image regeneration is improved, but complexity of data processing increases
Solution Approach 1:
The system dynamically determines the number and precision of disparity vectors needed for each tile based on local image content characteristics. Rather than using a fixed high-precision representation everywhere, the complexity of disparity data varies adaptively across different regions of the image.
Solution Approach 2:
Different regions of the image are processed with different levels of disparity precision. Tiles containing simple regions use fewer or coarser disparity vectors, while tiles with complex content use more or finer disparity vectors, optimizing the balance between fidelity and processing complexity locally.
Data Source
AI summary
A method for compressing data comprising a stereoscopic pair of images which provide parallax, when viewed by left and right eyes of an observer, involves generating (S31) the stereoscopic pair of images, each image comprising a plurality of display elements. A disparity value (S32) is generated indicating an amount by which a location of each display element in a first image of the stereoscopic pair of images is displaced compared to a displaced location of the same display element in a second image of the stereoscopic pair of images, such that a plurality of disparity values is generated. The number of different disparity values present in the plurality of disparity values is then reduced (S34) to produce a reduced set of different disparity values. A particular disparity value is determined from the reduced set of different disparity values to be associated with each of the locations and/or displaced locations of the display elements, and compressed data (S35) is generated comprising the first image of the stereoscopic pair of images, the reduced set of different disparity values, and information indicating the particular disparity value associated with each of the locations and/or displaced locations to be used to regenerate a regenerated second image of the stereoscopic pair of images corresponding to the second image.