Depth Map Filtering Using Texture and Depth Properties
Solution Overview
Problem
The high bitrate associated with 3D video content poses challenges for broadcasting due to limited bandwidth capacity of broadcasting infrastructures, and existing methods for depth map filtering do not effectively balance quality and bitrate for efficient compression and transfer of depth-enhanced multi-view content.
Innovation Solution
A method involving partitioning depth and texture images into blocks, determining property values for each block, and using these values to select filter parameters from a lookup table for filtering, which considers face detection and frequency domain characteristics to optimize filter strength and preserve foreground quality while reducing background detail.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If depth map filtering is applied to compress 3D video content, then bitrate is reduced, but image quality deteriorates
Solution Approach 1:
The patent applies different filter strengths to different regions of the depth map based on local image properties. Foreground regions with high texture complexity or face detections receive minimal filtering to preserve quality, while background regions with low texture complexity receive stronger filtering to reduce bitrate. This local differentiation resolves the contradiction by maintaining quality where needed while compressing where possible.
Solution Approach 2:
The patent dynamically adjusts filter parameters (filter strength, kernel size) based on depth property values and texture property values calculated for each block. By changing these parameters adaptively rather than applying a fixed filter, the system reduces bitrate in suitable regions while preserving quality in important regions, thus resolving the bitrate-quality contradiction.
2Device complexity
If uniform filtering is applied to the entire depth map, then processing complexity is reduced, but quality degradation occurs in important regions
Solution Approach 1:
The patent divides the depth map into blocks and calculates local properties (depth property value, texture property value) for each block. Based on these local properties, different filter parameters are selected from a lookup table for each block. This local adaptation maintains foreground quality while only moderately increasing processing complexity compared to uniform filtering.
Solution Approach 2:
The patent performs preliminary analysis of each block to determine depth property values and texture property values before applying filtering. Face detection and texture complexity assessment are done in advance, allowing the system to pre-determine appropriate filter parameters for each region. This preliminary action enables quality-preserving filtering in important regions while keeping overall complexity manageable.
3Productivity
If strong filtering is applied to reduce bitrate, then compression efficiency is improved, but detail loss increases
Solution Approach 1:
The patent uses a lookup table that maps combinations of depth property values and texture property values to appropriate filter parameters. This allows dynamic adjustment of filter strength based on local content characteristics, achieving good compression efficiency in suitable regions while preserving details in regions where strong filtering would cause unacceptable information loss.
Solution Approach 2:
The system uses calculated depth property values and texture property values as feedback to select appropriate filter parameters for each block. This feedback mechanism ensures that filtering strength is adapted to local content requirements, improving compression efficiency where possible while preventing detail loss in important regions through intelligent parameter selection.
Data Source
Figure 1A~1B
Figure 2~3
Figure 4
AI summary
There is inter alia a method comprising: partitioning a depth map image into at least one block; partitioning a texture image into at least one block, wherein the size and position of the at least one block of the texture image corresponds to the size and position of the at least one block of the depth image; determining a depth property value for the at least one block of the depth map image; determining a texture property value for the at least one block of the texture image; using the depth property value for the at least one block of the depth map image and the texture property value for the at least one block of the texture image to select a set of filter parameters from a look up table; and filtering the at least one block of the depth map image with a filter based on the selected set of filter parameters.