3D Video ROI Encoding via Depth-Based Disparity Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing region of interest (ROI) coding in 3D video coding may diminish perceptual quality if the selected criteria do not align with human interests, leading to discomfort during stereoscopic viewing.
Innovation Solution
Identifying disparities between multiple views in an image to determine the region of interest, which is then encoded at lower quantization compared to the remainder of the image, ensuring that the encoded image's quality aligns with human interests and enhances the 3D viewing experience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If ROI is selected based on motion and texture criteria, then encoding efficiency is improved, but perceptual quality diminishes when criteria do not align with human interests
Solution Approach 1:
The patent changes the selection parameter for ROI from traditional motion/texture criteria to depth-based criteria. By using depth information to identify regions that correspond to objects at different distances from the viewer, the encoding system aligns ROI selection with human visual interest, which naturally focuses on nearby objects. This parameter change resolves the contradiction by making ROI selection simultaneously efficient and perceptually accurate.
Solution Approach 2:
The patent introduces depth information as an intermediary between the encoding system and human visual perception. Depth maps serve as a mediator that translates three-dimensional spatial relationships into two-dimensional image regions, enabling the encoder to identify ROIs that correspond to objects at various depths. This intermediary allows the system to select ROIs that both optimize encoding efficiency and match human viewing interests.
2Device complexity
If uniform quantization is applied across the entire image, then encoding simplicity is maintained, but bitrate efficiency decreases when high quality is needed for all regions
Solution Approach 1:
The patent applies different quantization qualities to different regions of the image based on depth information. Regions corresponding to objects at closer depths (which are more likely to be of human interest) receive lower quantization (higher quality), while regions at farther depths receive higher quantization (lower quality). This local differentiation resolves the contradiction by maintaining encoding simplicity through automated depth-based region classification while improving bitrate efficiency through selective quality allocation.
3Manufacturing precision
If ROI is encoded at lower quantization, then perceptual quality of important regions is improved, but overall bitrate increases
Solution Approach 1:
The patent changes the basis for ROI selection from traditional criteria (motion, texture) to depth-based criteria. This parameter change enables more accurate identification of regions that truly warrant high-quality encoding, thereby improving perceptual quality of important regions while minimizing unnecessary bitrate consumption on less important regions.
Solution Approach 2:
The patent applies lower quantization (higher quality) only to specific depth-based regions rather than uniformly across the entire image. By confining high-quality encoding to ROI regions identified through depth information, the system improves perceptual quality where it matters most while controlling overall bitrate through higher quantization in non-ROI regions.
Data Source
AI summary
For coding at least one region of interest within an image of multiple views, disparities are identified between the multiple views. In response to the disparities, the at least one region of interest is identified. The at least one region of interest is encoded at lower quantization relative to a remainder of the image. The remainder of the image is encoded at higher quantization relative to the at least one region of interest.