TOF Phase Denoising With Dual Bilateral Filters for Depth Sensing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing TOF sensors face challenges in balancing depth disambiguation and determination while maintaining signal-to-noise ratio (SNR) and resolution, often requiring large pixels or extensive pixel binning, which compromises image quality.
Innovation Solution
A TOF sensor system that performs phase data denoising using two different strengths of filters, applying bilateral filtering to preserve resolution and improve SNR by utilizing multiple measurements and phase calculation blocks to generate high-fidelity and low-fidelity denoised phase data streams.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single strong filter is used for phase data denoising, then noise is reduced and SNR is improved, but depth disambiguation accuracy deteriorates due to excessive smoothing
Solution Approach 1:
The patent divides the phase data denoising process into two separate filtering operations with different strengths. A first filter with lower strength is applied to preserve depth disambiguation accuracy, while a second filter with higher strength is applied to improve SNR. This segmentation allows each filter to optimize for its specific purpose without compromising the other metric.
Solution Approach 2:
The patent applies different filtering strengths to different aspects of phase data processing. The first filter operates with parameters optimized for depth disambiguation, while the second filter uses parameters optimized for noise reduction. This local quality approach ensures that each filtering operation targets its specific goal with appropriate intensity.
2Reliability
If a single strong filter is used for phase data denoising, then noise is reduced, but resolution is lost due to excessive smoothing
Solution Approach 1:
The patent segments the filtering process into two distinct stages: first applying a mild filter that preserves resolution while removing obvious noise, then applying a stronger filter to the already-denised data to further reduce noise without severely impacting resolution. This staged approach achieves better overall noise reduction than a single strong filter applied directly to raw phase data.
3Measurement precision
If complex phase data denoising is performed, then depth determination accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies the first filter as a preliminary denoising step before depth disambiguation, and the second filter as a preliminary step before depth determination. By performing denoising in advance with appropriately tuned filter strengths, the subsequent depth calculation operations work with cleaner data, improving accuracy without requiring excessively complex real-time processing.
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
The system effectively enhances depth disambiguation and determination by maintaining resolution and reducing noise, optimizing SNR without sacrificing image quality.
Implementation Method 1
The active pixels in the array may include photosensitive elements such as pinned photodiodes that convert the incoming light into electric charge.
Implementation Method 2
The illumination module emits light onto an image scene having one or more objects. The emitted light reflects off of the one or more objects and is received by pixels in the sensor module.
Implementation Method 3
Based on the received light (e.g., the generated electrical charge), the sensor module can perform time-of-flight sensing calculations or operations to determine depth and other scene information.
Data Source
AI summary
A time-of-flight (TOF) sensing system may include an illumination module and a sensor module. The sensor module may include an array of sensor pixels, each sensor pixel configured to perform multiple measurements to generate corresponding pixel data for a TOF sensing operation. Signal processing circuitry may generate phase data based on the pixel data. Phase denoise circuitry in the signal processing circuitry may perform different types of filtering operations on the phase data such as perform two bilateral filters of varying strengths. The lower-fidelity denoised phase data may be used for depth disambiguation, while the higher-fidelity denoised phase data may be used for depth calculation. If desired, the phase denoise circuitry may perform averaging operations for one or both of these filtering operations using a Cartesian coordinate representation and efficiently using piecewise linear trigonometric approximations.