Feature Descriptor Sampling for Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating feature descriptors for image analysis in computer vision systems are inefficient, particularly in terms of processing cost and memory bandwidth, due to the need for separate sampling processes to determine rotation and extract feature descriptors.
Innovation Solution
A computer-implemented method and system that determines a set of samples characterizing a location in an image by sampling scale-space data, and generates a feature descriptor based on these samples, while also accounting for the measure of rotation to produce a rotation-invariant descriptor.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If separate sampling processes are used to determine rotation and extract feature descriptors, then feature matching accuracy is maintained, but processing cost and memory bandwidth increase
Solution Approach 1:
The patent combines the rotation determination sampling process and the feature descriptor extraction sampling process into a single unified sampling process. By sampling scale-space data at multiple length scales simultaneously and using the same samples for both rotation measurement and descriptor generation, the system eliminates redundant memory accesses and computational operations, thereby reducing processing cost and energy consumption while maintaining feature matching accuracy
Solution Approach 2:
The sampled scale-space data serves multiple functions simultaneously: it is used to determine the measure of rotation and to extract the feature descriptor. This multi-functional use of the sampling process eliminates the need for separate sampling operations, reducing overall processing cost and memory bandwidth requirements while preserving the accuracy of both rotation determination and feature matching
2Measurement precision
If separate sampling processes are used to determine rotation and extract feature descriptors, then feature matching accuracy is maintained, but memory bandwidth increases
Solution Approach 1:
The patent merges the memory access operations for rotation determination and feature descriptor extraction into a single sampling process. By retrieving scale-space data at multiple length scales once and reusing these samples for both purposes, the system significantly reduces total memory bandwidth consumption compared to making separate memory accesses for each function
Solution Approach 2:
The system performs preliminary sampling of scale-space data at multiple length scales before the main processing steps. This preliminary action prepares the samples in advance, allowing both rotation determination and feature descriptor extraction to proceed using the same pre-fetched data, thereby reducing overall memory bandwidth requirements
Data Source
AI summary
A computer-implemented method for generating a feature descriptor for a location in an image for use in performing descriptor matching in analysing the image, the method comprising determining a set of samples characterising a location in an image by sampling scale-space data representative of the image, the scale-space data comprising data representative of the image at a plurality of length scales; and generating a feature descriptor in dependence on the determined set of samples.