Depth Data Edge Detection for 3D Image Artifact Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 3D display devices, particularly auto-stereoscopic ones, suffer from resolution loss and require the viewer to remain at a fixed position, with existing methods for converting 2D image data to 3D data often resulting in visible artifacts due to de-occlusion, especially when dealing with textured backgrounds.
Innovation Solution
The method involves computing edges in depth-related data, measuring pixel value variation in background regions, and scaling depth data to reduce artifacts by shifting pixel values, thereby minimizing the visibility of holes created during the rendering process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If depth image based rendering is used to generate 3D images from 2D images, then 3D visualization capability is improved, but visible artifacts caused by de-occlusion increase
Solution Approach 1:
The patent applies local quality by differentiating the treatment of background regions based on texture complexity. The measure of variation is computed specifically for background pixels neighboring depth edges, and scaling is applied selectively to depth values in these regions. This localized approach preserves global 3D visualization capability while specifically targeting artifact reduction in problematic areas with high texture variation.
Solution Approach 2:
The patent changes the depth parameter by scaling depth values based on the measured variation in background pixel values. The scaling factor is derived from the measure of variation, which quantifies texture complexity. By dynamically adjusting the depth parameter according to local texture characteristics, the patent reduces artifacts in high-variation regions while maintaining depth accuracy in low-variation regions.
2Ease of operation
If auto-stereoscopic display devices are used to enable glasses-free 3D viewing, then viewer comfort is improved, but viewer position flexibility deteriorates
Solution Approach 1:
The patent applies dynamics by making the depth rendering adaptive to the viewer's position through the measure of variation calculation. The system dynamically adjusts the scaling of depth values based on the detected texture complexity in background regions, allowing the display to optimize image quality for different viewing conditions and positions, thereby improving both comfort and position flexibility.
3Object-affected harmful factors
If depth values are scaled to reduce artifacts in high variation regions, then artifact visibility is reduced, but depth accuracy is compromised
Solution Approach 1:
The patent applies local quality by selectively scaling depth values only in regions where the measure of variation exceeds a threshold. This localized scaling preserves depth accuracy in regions with low texture variation while reducing artifacts in regions with high texture variation, thus maintaining overall depth precision while eliminating visible artifacts.
Solution Approach 2:
The patent changes the depth parameter dynamically based on the measured variation. The scaling factor is computed as a function of the measure of variation, creating a continuous relationship between texture complexity and depth scaling. This adaptive parameter change reduces artifacts in high-variation regions while maintaining accurate depth representation in low-variation regions.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method of rendering an output image (202) on basis of an input image (200) and a corresponding matrix of depth related data (204), the input image (200) corresponding to a first viewpoint and the output image (202) corresponding to a second viewpoint being different from the first viewpoint, is disclosed. The method comprises: computing edges (212,312) in the matrix of depth related data (204), by computing derivatives of the depth related data in a predetermined direction (X); computing a measure of variation in pixel values in regions of the input image (200) corresponding to neighborhoods of the edges (212,312), the neighborhoods located at the background side of the edges (212,312); and computing the output image (202) by shifting respective pixel values of the input image (200) in the predetermined direction (X) on basis of scaled depth related data, wherein scaling is such that a relative large measure of variation results in a relative large reduction of depth.