3D Stereoscopic Image Compression via Global Motion Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing compression techniques for stereoscopic images do not consider their unique characteristics, leading to low compression efficiency and increased computational load in mobile communication terminals, which are limited by the larger amount of image information required for processing.
Innovation Solution
An apparatus and method that compress 3D stereoscopic images by computing global motion between two input images, using a difference estimator to find an overlapped image with the smallest disparity and encoding the residual image and position difference, thereby reducing computational and transmission loads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing compression techniques (MPEG4, JPEG) are used for stereoscopic images, then the images can be compressed, but compression efficiency is very low due to not considering stereoscopic image characteristics
Solution Approach 1:
The patent changes the processing parameters by treating stereoscopic images as two separate images (left and right views) with different disparities. The difference estimator computes position differences between corresponding blocks in the two images, and the encoder processes them differently based on their disparity values, optimizing compression for the specific characteristics of stereoscopic content
Solution Approach 2:
The patent segments the stereoscopic image into two separate image streams (first image and second image) and processes them independently through the difference estimator and encoder. This segmentation allows each image to be compressed considering its specific characteristics rather than treating them as a single complex structure
2Reliability
If stereoscopic images are transmitted or displayed, then three-dimensional visual information is provided, but the device requires more processing capacity since stereoscopic images have a larger amount of image information than still images
Solution Approach 1:
The patent extracts the essential visual information from stereoscopic images by computing only the necessary difference data between the two images. Instead of transmitting or processing the full high-resolution stereoscopic pair, the system extracts position differences and residual information, significantly reducing processing capacity requirements while maintaining the three-dimensional visual effect
Solution Approach 2:
The patent applies partial action by processing only the necessary portions of the stereoscopic images. The difference estimator compares blocks and computes position differences only where needed, and the encoder processes residual information selectively, avoiding excessive processing of all image data while still providing high-quality 3D visualization
3Ease of operation
If the amount of image information in stereoscopic images is reduced for mobile communication terminals, then processing limitations are addressed, but transmission traffic and storage requirements remain problematic
Solution Approach 1:
The patent creates a compressed representation (copy) of the stereoscopic image data by encoding position differences and residual information instead of the full image data. This copied representation requires less storage space and transmission bandwidth while allowing the terminal to reconstruct and display the stereoscopic images with full visual quality
Data Source
AI summary
An apparatus and method for efficiently compressing three-dimensional stereoscopic images are provided. The apparatus and method include a search for Global motion between left and right images input through two image input units mounted in a mobile communication terminal. The images are arranged and a difference between the two images is computed. The images are compressed using the difference between the two images and a reference image. Accordingly, because the left and right images are similar, a computation amount can be reduced since the global motion is computed only once.