Depth-Dependent Spatial Filtering for Multi-View Display Crosstalk
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-view displays face issues with image overlap and crosstalk, leading to ghost images and aliasing effects due to a limited number of view-directions, which affect the perceived image quality.
Innovation Solution
A method of rendering image data using a depth-dependent spatial filter that maps image elements from a first image to a second image, with the filter strength determined by the difference in depth, allowing for efficient reduction of artefacts like crosstalk and aliasing by filtering in the input domain rather than the output domain, and incorporating band-pass and high-pass filtering techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If images for different view-directions are presented in a multi-view display, then the viewer can experience motion parallax and stereoscopic cues, but image overlap occurs leading to ghost images and cross-talk between images
Solution Approach 1:
The patent applies spatial filtering to the input image domain before the image is rendered for multi-view display. This preliminary action pre-processes the image data to reduce cross-talk and ghosting artifacts before they occur during the multi-view rendering process, rather than attempting to correct them after they appear in the output
Solution Approach 2:
The patent applies different filtering strengths to different regions of the image based on depth information. Objects at different depths receive different levels of spatial filtering, with the filter strength determined by the depth of the image element relative to a reference depth plane. This local differentiation allows the system to maintain image quality for objects at the reference depth while reducing cross-talk for objects at other depths
2Ease of manufacture
If the number of view-directions is limited to eight or nine, then the display can be implemented with current technology, but aliasing effects occur in some view-directions
Solution Approach 1:
The patent performs spatial filtering in the input domain before the image is sampled into multiple view-directions. This preliminary filtering prevents aliasing artifacts from occurring during the sampling process, allowing the system to use a limited number of view-directions (eight or nine) without suffering from aliasing effects that would normally result from such coarse sampling
Solution Approach 2:
The patent changes the spatial frequency content of the input image through filtering operations before rendering. By modifying the frequency domain characteristics of the image data in advance, the system can accommodate the limited angular resolution of current multi-view displays without exhibiting visible aliasing artifacts
3Object-affected harmful factors
If spatial filtering is performed in the output domain on multiple images, then cross-talk can be reduced, but the processing complexity and computational cost increase significantly
Solution Approach 1:
The patent inverts the conventional approach by performing spatial filtering in the input domain rather than the output domain. Instead of filtering multiple rendered images after they have been generated for different view-directions, the system applies a single spatial filter to the original input image before rendering, which achieves cross-talk reduction with much lower computational complexity
Solution Approach 2:
The patent merges the filtering operation into a single step applied to the input image, rather than requiring separate filtering operations for each view-direction in the output domain. This consolidation reduces the total number of filtering operations from multiple (one per view) to a single operation, significantly reducing processing complexity
Data Source
Figure 1~3
Figure 4~5C
Figure 6A~7B
AI summary
The invention pertains to rendering of image data for a multi-view display, such as image data for a lenticular auto-stereoscopic display. The method comprising the steps of receiving an image signal representing a first image, the first image comprising 3D image data, spatially filtering the first image signal to provide a second image signal, the second image signal representing a second image, the spatial filtering being such as a low- pass filter, a high-pass filter or a combination of a low-pass and a high-pass filter, a strength of the spatial filter is determined by a reference depth of the first image and a depth of an image element of the first image, and sample the second image to a plurality of sub-images, each sub-image being associated with a view direction of the image.