Adaptive Window Sizing for Optical Flow Calculation Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing devices calculate optical flow with uniform window sizes, leading to increased errors when objects at different distances and positions are within the same window, resulting in inaccurate and potentially accumulated error.
Innovation Solution
An image processing device that adjusts window sizes based on the position and relative distance of objects within an image, using position information from sensors like radars to set appropriate window sizes for accurate optical flow calculation, reducing the impact of background pixels and improving calculation accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a uniform window size is used for calculating optical flow, then the calculation process is simple and consistent, but the optical flow accuracy decreases when objects at different distances are located within the same window
Solution Approach 1:
The patent applies local quality by setting different window sizes for different regions of the image based on object distance. Specifically, the image is divided into a central region and peripheral regions, with the central region using a first window size and peripheral regions using a second window size. This allows each region to have optimized window dimensions suited to the typical object sizes and motion characteristics in that area, thereby improving optical flow accuracy without uniformly increasing complexity across the entire image.
Solution Approach 2:
The patent implements dynamics by making the window size adaptive rather than fixed. The window size is dynamically adjusted based on the object's distance from the camera and its position within the image. Objects closer to the camera use larger windows, while distant objects use smaller windows. This dynamic adaptation allows the system to maintain high accuracy across varying scene depths while avoiding the computational burden of using uniformly large windows throughout the image.
2Area of stationary object
If a large window size is used to ensure objects are fully captured, then the object detection coverage is improved, but the optical flow calculation accuracy decreases due to inclusion of excessive background pixels
Solution Approach 1:
The patent applies local quality by setting different window sizes for different regions of the image based on object distance. Specifically, the image is divided into a central region and peripheral regions, with the central region using a first window size and peripheral regions using a second window size. This allows each region to have optimized window dimensions suited to the typical object sizes and motion characteristics in that area, thereby improving optical flow accuracy without uniformly increasing complexity across the entire image.
Solution Approach 2:
The patent changes the window size parameter based on object distance and position. By adjusting this critical parameter dynamically, the system ensures that the window is large enough to capture the entire object while minimizing inclusion of surrounding background. This parameter adaptation resolves the contradiction between coverage and precision by making the window size proportional to the expected object size at different depths.
3Measurement precision
If a small window size is used to reduce background pixel interference, then the optical flow calculation accuracy for distant objects improves, but the object may not be fully captured within the window
Solution Approach 1:
The patent applies local quality by setting different window sizes for different regions of the image based on object distance. Specifically, the image is divided into a central region and peripheral regions, with the central region using a first window size and peripheral regions using a second window size. This allows each region to have optimized window dimensions suited to the typical object sizes and motion characteristics in that area, thereby improving optical flow accuracy without uniformly increasing complexity across the entire image.
Solution Approach 2:
The patent changes the window size parameter based on object distance and position. By adjusting this critical parameter dynamically, the system ensures that the window is large enough to capture the entire object while minimizing inclusion of surrounding background. This parameter adaptation resolves the contradiction between coverage and precision by making the window size proportional to the expected object size at different depths.
4Ease of operation
If uniform window sizes are used for all objects, then the calculation method is consistent and simple, but the computational load increases when processing images with objects at various distances
Solution Approach 1:
The patent applies segmentation by dividing the image into distinct regions (central region and peripheral regions) and applying different window sizes to each region. This segmentation allows the system to process different parts of the image with appropriately sized windows, reducing the total number of calculations required compared to using uniformly large windows across the entire image, while maintaining systematic processing through defined regional rules.
Data Source
AI summary
An image processing device for improving the accuracy of optical flow calculation when an optical flow is calculated in a window unit. An image processing device for calculating an optical flow on the basis of image information within a window for a processing target using a plurality of images captured at different times includes position acquisition means which acquires position information of the processing target and setting means which sets a size of a window for calculating an optical flow on the basis of the position information acquired by the position acquisition means.