Adaptive Encoding for Fish-Eye Stereoscopic Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for encoding stereoscopic images captured with fish-eye lenses face challenges in maintaining image quality due to bandwidth limitations and distortion, leading to decreased perceived depth and quality, especially when capturing images with fish-eye lenses that introduce more distortions in certain portions of the image.
Innovation Solution
The method involves determining the depth of pixels in stereoscopic images and adjusting the number of bits per pixel based on depth levels, object size, activity level, and the portion of the fish-eye lens used for capture, allowing for adaptive encoding that prioritizes preserving edges and depth perception by using different bit allocations and block sizes for encoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If lossy encoding is used to reduce bandwidth consumption, then data transmission efficiency is improved, but image quality and perceived depth deteriorate
Solution Approach 1:
The patent applies local quality by allocating different bit depths to different spatial regions of the image based on depth information. Pixels corresponding to closer objects (which are more important for depth perception) are encoded with higher bit depth (e.g., 10-12 bits), while pixels corresponding to distant objects are encoded with lower bit depth (e.g., 8 bits). This resolves the contradiction by preserving critical depth information where needed while reducing overall bandwidth consumption.
Solution Approach 2:
The patent segments the image into multiple depth layers or regions based on depth mapping. Each segment is then independently encoded with appropriate bit depth and compression parameters. This segmentation allows selective preservation of depth-critical regions while applying more aggressive compression to less important regions, thereby maintaining perceived depth while improving transmission efficiency.
2Ease of operation
If uniform bit allocation is used for all pixels, then encoding simplicity is maintained, but depth perception quality deteriorates due to insufficient bits for close objects
Solution Approach 1:
The patent implements local quality by determining depth for each pixel or block of pixels and allocating bits dynamically based on that depth information. Close objects receive higher bit allocation to preserve edges and depth cues, while distant objects receive lower bit allocation. This maintains encoding simplicity through automated depth-based rules while significantly improving depth perception quality.
3Area of stationary object
If fish-eye lenses are used to capture wide field of view, then scene coverage is improved, but image distortion increases in certain portions
Solution Approach 1:
The patent addresses fish-eye distortion by applying different encoding parameters to different regions of the image. Central regions with less distortion receive standard encoding, while peripheral regions with higher distortion receive adjusted encoding parameters such as lower bit depth or different prediction block sizes. This local adaptation maintains the wide field of view benefit while minimizing the visual impact of distortion in problematic areas.
Data Source
AI summary
A method and system for encoding a stereoscopic image pair is disclosed. Groups of pixels are analyzed to determine the depth of each pixel group. The number of bits per pixel used to encode each pixel group is selected based on the depth of that pixel group. Therefore, images of objects closer to the camera pair, which appear closer to the viewer, are encoded with a larger number of bits per pixel than objects perceived to be farther from the viewer. The number of bits per pixel may also be increased based on a number of objects depicted or motion detected. The size of prediction blocks used to encode image portions may also be determined based on an angular distance of an image portion relative to the center of the frame. Therefore, smaller prediction blocks may be used to encode image portions closer to the center of the frame.