Ray-Space Free View Generation via Radon Transform
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Ray-Space interpolation techniques produce numerous matching errors and do not provide acceptable rendered views without advance knowledge of the scene, making it difficult to generate arbitrary virtual views in applications like Free Viewpoint Television (FTV).
Innovation Solution
The proposed architecture employs a Radon transform to extract feature points from an epipolar plane image (EPI), construct a feature epipolar plane image (FEPI), and apply techniques such as intensity functions and local maximum peak constraints to determine the slope of lines, thereby reducing the number of candidate interpolation slopes and improving the accuracy of Ray-Space interpolation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional Ray-Space interpolation techniques are used, then the process can be performed without advance knowledge of the scene, but numerous matching errors are produced and rendered views are not acceptable
Solution Approach 1:
The patent applies preliminary action by pre-processing the EPI to extract feature points and construct the FEPI before performing the Radon transform. This preliminary extraction of meaningful features (corners, edges, lines) from the EPI enables the subsequent interpolation to be guided by semantically significant structures rather than raw pixel data, thereby improving accuracy without requiring advance knowledge of the 3D scene.
Solution Approach 2:
The patent extracts feature points from the EPI and represents them as lines in the FEPI, separating the essential geometric information from the redundant pixel data. By taking out only the meaningful features (corners, edges, lines) and representing them as simplified line structures in the FEPI, the system reduces the complexity of the interpolation task while maintaining accuracy in generating virtual views.
2Device complexity
If Ray-Space interpolation is performed without limiting candidate slopes, then the process is simpler, but many incorrect slopes are considered reducing interpolation accuracy
Solution Approach 1:
The patent segments the set of possible interpolation slopes into two distinct groups: those that correspond to actual scene structures (correct slopes) and those that do not (incorrect slopes). By segmenting the candidate slopes based on whether they align with extracted feature lines in the FEPI, the system can efficiently filter out incorrect slopes while retaining the correct ones, maintaining simplicity while improving precision.
Solution Approach 2:
The patent implements feedback by using the extracted feature points and their represented lines in the FEPI to guide and constrain the interpolation process. The FEPI acts as a feedback mechanism that provides geometric constraints on which slopes are valid, allowing the interpolation algorithm to adjust its search based on the extracted feature information and thereby improve slope determination accuracy.
Data Source
AI summary
The claimed subject matter relates to an architecture that can facilitate more efficient free view generation in Ray-Space by way of a Radon transform. The architecture can render virtual views based upon original image data by employing Ray-Space interpolation techniques. In particular, the architecture can apply the Radon transform to a feature epipolar plane image (FEPI) to extract more suitable slope or direction candidates. In addition, the architecture can facilitate improved block-based matching techniques in order to determine an optimal linear interpretation direction.