Multi-Camera Depth Image Encoding for Lower 3D Transmission Load
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increase in the number of viewpoints for generating depth images to enhance the reproducibility of a 3D model leads to an increase in data transmission, which can compress the transmission band.
Innovation Solution
An image processing apparatus that identifies a representative camera for each partial region of a subject from multiple depth images, generates changed depth and texture images by reducing pixel value differences, and encodes these images to reduce data transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the number of viewpoints of depth images is increased to enhance reproducibility of 3D model, then the quality of 3D model restoration is improved, but the amount of data to be transmitted increases and transmission band is compressed
Solution Approach 1:
The patent merges redundant depth information from multiple viewpoint images by identifying corresponding partial regions and eliminating duplicate depth data. This combines multiple sources of information into a single optimized depth map that retains all necessary 3D reconstruction data while removing redundancy, thereby reducing transmission data volume without compromising 3D model restoration quality
Solution Approach 2:
The patent extracts only the essential depth information needed for 3D model restoration by identifying and removing redundant depth data from multiple viewpoint images. By extracting and eliminating duplicate depth values from corresponding partial regions across different viewpoints, the system transmits only the minimum necessary data while preserving complete 3D reconstruction capability
2Reliability
If the number of viewpoints of depth images is increased to enhance reproducibility of 3D model, then the quality of 3D model restoration is improved, but the transmission band is compressed
Solution Approach 1:
The patent merges redundant depth information from multiple viewpoint images by identifying corresponding partial regions and eliminating duplicate depth data. This combines multiple sources of information into a single optimized depth map that retains all necessary 3D reconstruction data while removing redundancy, thereby reducing transmission data volume without compromising 3D model restoration quality
Solution Approach 2:
The patent segments the depth images into multiple partial regions and processes each region independently by identifying corresponding regions across different viewpoints. This segmentation allows for efficient identification and elimination of redundant depth information in each region, simplifying the overall transmission data structure while maintaining complete 3D reconstruction capability
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A first image processing apparatus 20 acquire a depth image including a depth value determined based on a distance between a subject and a camera, for each of a plurality of cameras. Next, a determination unit 304 identifies, based on a first partial region of the subject that is included in a specific depth image among the plurality of depth images a second partial region corresponding to the first partial region, from a different depth image. A changed depth image generation unit 305 generate a changed depth image signal-processed in such a manner as to reduce a difference between depth values of the second partial region of the different depth image. An encoding unit 307 encodes a plurality of the generated changed depth images, and a transmission unit 308 transmits the encoded changed depth images.