Stereoscopic Video Quality Prediction via Image Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current technologies lack a full reference system for predicting the subjective quality of stereoscopic video, which can cause visual discomfort due to horizontal disparity between left and right images, and existing methods are not scalable or accurate for 3D video quality assessment.
Innovation Solution
A method that compares and combines the left and right sub-components of a 3D image to generate a predictive quality rating, using disparity measurements to fuse the images into a 2D representation for analysis, and includes a quality prediction system that compares test and reference stereoscopic videos to produce a DMOS rating.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If left and right 2-D images are measured separately and averaged, then quality assessment can be performed using existing 2-D techniques, but the result does not accurately predict subjective quality of stereoscopic video
Solution Approach 1:
The patent introduces a fuser as an intermediary component that combines left and right images into a fused image, which then serves as input for quality assessment. This mediator transforms the two separate 2-D images into a single representation that captures the stereoscopic quality characteristics, enabling accurate prediction of subjective quality while using existing 2-D assessment tools.
Solution Approach 2:
The patent transitions from assessing two separate 2-D images to creating a fused image that represents the stereoscopic combination. This dimensionality change allows the system to move from evaluating individual views to evaluating the combined stereoscopic experience, thereby improving prediction accuracy for 3-D video quality.
2Measurement precision
If a full reference system for stereoscopic video quality prediction is developed, then accurate subjective quality prediction is achieved, but system complexity increases compared to existing 2-D methods
Solution Approach 1:
The patent segments the quality assessment process into distinct functional modules: a disparity measurement module that calculates horizontal disparities, a fuser module that combines left and right images using disparity data, and a picture quality analyzer that evaluates the fused image. This segmentation allows each component to perform its specific function efficiently, managing system complexity through modular design.
Solution Approach 2:
The patent creates a system that can handle both stereoscopic and 2-D video quality assessment. The fuser can process stereoscopic pairs to create fused images for 3-D assessment, while also being capable of handling conventional 2-D video through appropriate configuration, thereby providing universal functionality across different video types.
3Adaptability or versatility
If 3-D video is generated from existing 2-D video or repurposed for different formats, then content reusability is improved, but quality control and assurance become more difficult
Solution Approach 1:
The patent performs disparity measurement and image fusion as preliminary actions before quality assessment. By pre-calculating disparity maps and fusing images in advance, the system prepares the stereoscopic content for evaluation, ensuring that quality control is integrated into the content creation and repurposing workflow rather than being an afterthought.
Data Source
AI summary
A method of generating a predictive picture quality rating makes a disparity measurement of a three-dimensional image by comparing left and right sub-components of the three-dimensional image. Then the left and right sub-components of the three-dimensional image are combined (fused) into a two-dimensional image, using data from the disparity measurement for the combination. A predictive quality measurement is then generated based on the two-dimensional image, and further including quality information about the comparison of the original three-dimensional image.