Adaptive Depth-Map Filtering for Spurious Pixel Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods, such as the bilateral filter, are inefficient in removing spurious pixels from depth-maps as they preserve sharp edges but fail to effectively filter out spurious pixels, and increasing the parameter σd would also dissolve sharp edges.
Innovation Solution
A computer-implemented method that determines coefficients for multiple weight functions based on geometrical distance and intensity, selecting the weight function with the maximum sum of coefficients to filter spurious pixels while preserving sharp edges, using minimum or maximum filters, and generating convolution filters from these weight functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Shape
If a bi-lateral filter is used to preserve sharp edges, then edge preservation is improved, but spurious pixels are not effectively removed
Solution Approach 1:
The patent applies dynamics by making the filtering process adaptive rather than static. Multiple weight functions are computed with different characteristics, and the system dynamically selects the most appropriate weight function for each pixel based on local depth-map characteristics. This allows the filter to adapt its behavior to preserve edges in some regions while removing spurious pixels in others.
Solution Approach 2:
The patent changes parameters by computing multiple weight functions with different parameter settings (different σd and σr values) and selecting the most appropriate one for each pixel. This parameter variation enables the filter to adjust its strength and behavior locally, achieving both edge preservation and spurious pixel removal in different regions of the depth-map.
2Reliability
If the parameter σd is increased to remove spurious pixels, then spurious pixel removal is improved, but sharp edges are dissolved
Solution Approach 1:
The patent applies local quality by computing multiple weight functions with different local characteristics and selecting the most appropriate one for each pixel's neighborhood. This allows different filtering strengths to be applied locally - stronger filtering in regions with spurious pixels and weaker filtering in regions with important edge details, achieving both goals simultaneously in different locations.
Solution Approach 2:
The system dynamically selects among multiple pre-computed weight functions based on local characteristics, effectively making the filtering strength adaptive rather than uniform. This dynamic selection allows the system to apply stronger filtering (higher σd) where needed to remove spurious pixels while maintaining weaker filtering (lower σd) where edge preservation is critical.
3Reliability
If multiple weight functions are computed and selected based on maximum sum of coefficients, then filtering effectiveness is improved, but computational complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-computing multiple weight functions with different characteristics before the actual filtering process. These pre-computed weight functions are then simply selected based on the maximum sum of coefficients criterion, avoiding the need for complex real-time computations during filtering. This preliminary preparation reduces the computational burden during the actual filtering operation.
Solution Approach 2:
The patent segments the filtering process into distinct stages: computing multiple weight functions with different characteristics, selecting the most appropriate one based on the maximum sum criterion, and applying the selected weight function. This segmentation allows each stage to be optimized independently and simplifies the overall computational complexity by breaking down the complex filtering task into manageable steps.
Data Source
Figure 1~3
Figure 2
AI summary
The invention concerns a method for filtering spurious pixels in a depth-map. It is important to filter a depth-map to remove the spurious pixels. One issue when filtering spurious pixels, is to remove them while preserving the boundaries of the real details of the depth-map. A known approach to filter spurious pixels is to use a bi-lateral filter. The bi-lateral filter offers an efficient filtering which enables to preserve the sharp edges of the depth-map while smoothing faint details or noise. However, the bi-lateral filter is not efficient to remove the spurious pixels. Indeed, the spurious pixels are showing sharp edges versus the pixels directly surrounding them. In other words, the bi-lateral filer conserves the spurious pixels even after several iterations of filtering. The proposed solution enables to remove spurious pixels in a depth-map while preserving the sharp edges of the depth-map and smoothing faint details or noise.