Depth Edge Preserving Filter for Time-of-Flight Depth Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Time-of-flight depth imaging technologies face challenges in effectively preserving depth edges while mitigating noise, as existing filters often distort depth maps and fail to maintain consistent filtering weights across images, leading to inaccurate distance estimates.
Innovation Solution
The development of depth edge preserving filters inspired by bilateral and guided filters, which calculate filter strengths based on depth variance to detect edges and maintain consistent weights across storage units, ensuring accurate depth edge detection and noise reduction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing filters are applied to ToF depth images, then noise is reduced, but depth edges are distorted and filtering weights are inconsistent across images
Solution Approach 1:
The patent applies different filtering strengths to different regions of the depth image based on local depth variance. Areas with high depth variance (edges) receive weaker filtering to preserve edge sharpness, while areas with low depth variance (smooth regions) receive stronger filtering to reduce noise. This local adaptation resolves the contradiction by making the filter behavior spatially variable rather than uniform.
Solution Approach 2:
The filtering strength is dynamically adjusted based on the calculated depth variance at each location. The filter automatically adapts its parameters according to the local characteristics of the depth image, transitioning between strong and weak filtering modes depending on whether the region is identified as an edge or smooth area. This dynamic behavior enables simultaneous noise reduction and edge preservation.
2Measurement precision
If strong filtering is applied to reduce noise, then noise in depth estimates is reduced, but depth edge sharpness is lost
Solution Approach 1:
The filter applies different filtering intensities to different spatial locations based on local depth variance calculations. Smooth regions with low variance receive strong filtering for noise reduction, while edge regions with high variance receive weak filtering to maintain edge sharpness. This resolves the contradiction by making filtering strength location-dependent rather than uniform across the entire image.
Solution Approach 2:
The filtering parameters (particularly the filtering strength) are changed based on the calculated depth variance. The system dynamically adjusts the filter parameters according to local image characteristics, enabling the filter to transition between aggressive noise reduction in smooth areas and conservative edge preservation in high-variant areas, thus resolving the trade-off between noise reduction and edge sharpness.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
These filters effectively preserve depth edges, reduce noise in depth estimates, and provide a confidence map for reliable depth measurements, enhancing the accuracy and reliability of time-of-flight depth imaging systems.
Implementation Method 1
The TOF image sensor is further equipped with a light source that illuminates objects whose distances from the device are to be measured by detecting the time it takes the emitted light to return to the image sensor
Implementation Method 2
The measurement of this time (i.e. the time of flight) can be used for a time standard (such as an atomic fountain), as a way to measure velocity or path length through a given medium, or as a way to learn about the particle or medium
Data Source
AI summary
Time of Flight (ToF) depth image processing methods. Depth edge preserving filters are disclosed with superior performance to standard edge preserving filters applied to depth maps. In particular, depth variance is estimated and used to filter while preserving depth edges. In doing so, filter strength is calculated which can be used as an edge detector. A confidence map is generated with low confidence at pixels straddling a depth edge, and which reflects the reliability of the depth measurement at each pixel.