Image Correlation Reaggregation for Precise Optical Mouse Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optical mouse tracking technologies suffer from inaccurate and distorted image correlation processing due to issues in macropixel and subpixel calculations, leading to unreliable tracking, especially at low speeds or shallow angles.
Innovation Solution
A reaggregation process is applied to the cross-correlation results, combining coincidental peaks with their neighbors and pre-distorting the distribution of values around the peak to correct for distortion, allowing for more accurate macropixel and subpixel displacement calculations without altering the existing algorithm.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If cross-correlation is performed on quantized image frames to determine displacement, then macropixel calculation is achieved, but the peak becomes disaggregated into a 2x2 plateau due to subpixel motion, reducing measurement precision
Solution Approach 1:
The patent segments the cross-correlation result into a 2x2 grid of pixels surrounding the peak location. By analyzing each pixel's contribution separately and applying different weighting factors based on their position relative to the peak, the method reconstructs the true peak magnitude that would otherwise be dispersed across the 2x2 plateau, thereby resolving the contradiction between precision and reliability.
Solution Approach 2:
The patent applies local quality by assigning different weights to different pixels within the 2x2 grid based on their proximity to the peak. Pixels closer to the peak center are weighted more heavily than those at the corners, allowing the method to reconstruct the true peak magnitude while accounting for the dispersed correlation values, thus improving both measurement precision and peak detection reliability.
2Measurement precision
If subpixel displacement calculation is performed using center of mass on the 2x2 array, then subpixel precision is achieved, but distortion and bias are introduced due to arbitrary offsets from adjacent features
Solution Approach 1:
The patent applies local quality by assigning different weights to different pixels within the 2x2 grid based on their proximity to the peak. Pixels closer to the peak center are weighted more heavily than those at the corners, allowing the method to reconstruct the true peak magnitude while accounting for the dispersed correlation values, thus improving both measurement precision and peak detection reliability.
Solution Approach 2:
The patent uses feedback by iteratively refining the peak location and magnitude estimates. The method adjusts the weighting factors and peak position based on the observed correlation pattern, continuously improving the accuracy of the subpixel displacement calculation while compensating for distortions introduced by adjacent features, thereby resolving the contradiction between precision and calculation accuracy.
3Measurement precision
If paraboloid fit is applied to 3x3 array values centered on global peak, then subpixel calculation is improved, but distortion accumulates errors at low speeds or shallow angles
Solution Approach 1:
The patent segments the analysis into two independent parts: first determining the macropixel displacement by identifying the peak location in the cross-correlation array, then calculating subpixel displacement separately using the reaggregated values. This segmentation allows each calculation to be optimized independently, improving overall reliability while maintaining precision even at low speeds or shallow angles where distortion accumulation would otherwise occur.
Data Source
AI summary
A method of image correlation includes: obtaining a first image frame with a set of distinct features; obtaining a second image frame at least partially overlapping the first image frame; carrying out a cross-correlation operation of the first and second image frames to obtain a cross-correlation result indicating correlation values for different offset positions of the second image frame with respect to the first image frame, the highest correlation value with its associated offset position in the cross-correlation result indicating the most likely offset position of the second image frame with respect to the first image frame; and carrying out a reaggregation operation of the cross-correlation result to generate a distinct global maximum correlation value from a plurality of non-distinct correlation values in an aggregated cross-correlation result. The reaggregation operation involves carrying out a convolution of the cross-correlation result with a 1×2, 2×1, or 2×2 array of 1's.