3D Depth Extraction Using Weighted Image Averaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D image acquisition methods, such as those using the time-of-flight (TOF) method, face challenges in accurately extracting depth information due to random noise from light sources, optical modulators, and image pickup devices, leading to errors and increased memory and computation requirements.
Innovation Solution
A method involving the sequential projection of multiple periodic waves of light onto an object, with corresponding optical modulation signals, and the use of pre-calculated primary and secondary weights to generate averaged images, allowing for the calculation of average phase delay and accurate depth determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple images are acquired and averaged to remove random noise, then measurement precision is improved, but device complexity and computation amount increase exponentially
Solution Approach 1:
The patent pre-calculates and stores weight values for combining multiple images before actual depth measurement. These pre-computed weights are stored in a lookup table, allowing the system to quickly retrieve and apply appropriate weights during operation without performing complex calculations in real-time, thus reducing computational complexity while maintaining noise removal effectiveness
Solution Approach 2:
The patent changes the approach from uniform averaging to weighted averaging based on signal strength parameters. By adjusting the weights dynamically according to the strength of the modulated signal in each image, the system optimizes noise reduction while adapting to varying lighting conditions and object reflectivity, achieving better measurement precision without proportional increase in computation
2Measurement precision
If multiple images are acquired and averaged to remove random noise, then measurement precision is improved, but memory usage increases exponentially
Solution Approach 1:
The patent performs preliminary calculations of weight values and stores them in a compressed lookup table structure. This pre-computed weight data is significantly smaller than storing all original image data, allowing the system to retrieve only the necessary weight information for each pixel during processing, thereby reducing memory requirements while enabling effective noise removal through weighted averaging
Solution Approach 2:
The patent extracts only the essential weight information from the full image data and separates it into a compact lookup table. This extraction allows the system to discard the bulk of the original image data after extracting the necessary weight values, significantly reducing the amount of data that needs to be stored in memory while maintaining the capability to perform accurate weighted averaging for noise removal
3Measurement precision
If a least square fitting method using pseudo-inverse matrix is used to average images, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent pre-calculates the weight values needed for image averaging and stores them in a lookup table before actual depth measurement. This eliminates the need to perform complex least square fitting calculations with pseudo-inverse matrices during real-time operation, significantly reducing algorithmic complexity while maintaining the ability to remove random noise through weighted averaging
Solution Approach 2:
The patent replaces the complex mathematical mechanism of least square fitting with a simpler weighted averaging approach using pre-computed weights. Instead of solving systems of linear equations through matrix inversion, the system simply retrieves and applies pre-calculated weight values from lookup tables, substituting a complex computational mechanism with a simpler data retrieval and multiplication process
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
This approach efficiently removes random noise and achieves accurate depth information extraction with reduced computational complexity and memory usage, enhancing the performance of 3D image acquisition systems.
Implementation Method 1
modulating N pieces of reflection light reflected from the object with an optical modulation signal having a gain waveform
Implementation Method 2
measuring a light flight time from when illumination light is irradiated onto an object to when light reflected from the object is received by a light receiver
Implementation Method 3
receiving light having the same wavelength as the specific wavelength, which is reflected from the object, by a light receiver
Data Source
AI summary
A 3 dimensional (3D) image acquisition apparatus and a method of extracting depth information in the 3D image acquisition apparatus are provided. The method of extracting depth information includes sequentially projecting N (N is a natural number equal to or greater than 3) different pieces of projection light onto a object; modulating N pieces of reflection light reflected from the object with an optical modulation signal having a gain waveform; generating N images by capturing the N pieces of modulated reflection light; generating a first averaged image by multiplying the N images by primary weights and generating a second averaged image by multiplying the N images by secondary weights; acquiring an average phase delay from the first and second averaged images; and calculating a distance to the object from the average phase delay.