Depth Map Generation for 2D to 3D Video Conversion

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional video coding techniques struggle with effectively converting two-dimensional (2D) video to three-dimensional (3D) video, particularly in scenarios where depth maps are required for view synthesis, as they often rely on intensive computations and may not accurately capture depth cues from monoscopic sources.

Innovation Solution

The proposed techniques involve a depth map initialization process where initial depth values are assigned to video units based on motion and adjusted based on color, with the generation of an initialized depth map that can be refined further, depending on scene changes and motion levels, to produce a final depth map for 2D to 3D video conversion.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional video coding techniques are used for 2D to 3D conversion, then the conversion process can be performed, but the computational complexity becomes excessive and depth accuracy is insufficient

Engineering Contradiction:
Improvedepth accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The depth map generation process is segmented into distinct stages: initialization phase (assigning initial depth values based on motion), refinement phase (adjusting depth values based on color), and finalization phase (applying scene change detection and motion level analysis). This segmentation allows each stage to focus on specific aspects of depth estimation, improving overall accuracy while managing computational load efficiently.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies preliminary action by performing motion-based depth initialization before color-based refinement. By pre-establishing depth values using motion information from previous frames, the system creates a preliminary depth map that captures temporal depth cues. This preliminary depth estimation is then refined using color information, reducing the computational burden of exhaustive color analysis while maintaining depth accuracy.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If intensive computations are applied to generate accurate depth maps, then depth accuracy improves, but processing time and computational resources increase significantly

Engineering Contradiction:
Improvedepth accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies partial action by selectively refining depth values based on color information only for pixels that require refinement, rather than processing all pixels uniformly. The system identifies pixels where motion-based depth estimation is insufficient and applies color-based refinement only to those specific regions. This selective approach maintains depth accuracy where needed while significantly reducing overall processing time and computational resource consumption.

Inventive Principle:
Principle #16Partial or excessive action

3Productivity

If motion-based depth initialization is used, then processing efficiency improves, but depth accuracy deteriorates for static regions

Engineering Contradiction:
Improveprocessing efficiencyVSAvoiddepth accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies local quality by differentiating depth estimation strategies based on local pixel characteristics. For pixels in motion regions, motion-based depth initialization is used to maintain processing efficiency. For pixels in static regions where motion cues are insufficient, color-based refinement is applied to improve depth accuracy. This localized adaptation of estimation strategies ensures high processing efficiency overall while maintaining depth accuracy in static regions where it matters most.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system implements feedback by using color information to verify and adjust motion-based depth estimates. After initial depth values are assigned based on motion, the refinement process provides feedback by analyzing color gradients and consistency with neighboring pixels. This feedback mechanism identifies and corrects depth estimation errors in static regions, improving overall depth accuracy while maintaining the efficiency benefits of motion-based initialization.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP2491722B1Depth map generation techniques for conversion of 2d video data to 3D video data
Publication Date: 2015.09.23 QUALCOMM INC
  • EP2491722B1 patent drawingFigure 1
  • EP2491722B1 patent drawingFigure 2
  • EP2491722B1 patent drawingFigure 3

AI summary

This disclosure describes techniques for generating depth maps for video units, such as video frames or slices video frames. The techniques may be performed by a video encoder in order to convert two-dimensional (2D) video to three-dimensional (3D) video. The techniques may alternatively be performed by a video decoder in order to convert received 2D video to 3D video. The techniques may use a combination of motion and color considerations in the depth map generation process.