Triangle Filter Keypoint Extraction Speed
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing keypoint extraction methods, such as SIFT and SURF, face high computational costs and inefficiencies in scale-space extraction and description, which hinder their performance in applications like 3D reconstruction and object recognition.
Innovation Solution
A novel method using piece-wise triangle filters to approximate the determinant of the Hessian, combined with an optimized DAISY-like descriptor that reuses computations and reduces sampling space, leading to faster and more precise keypoint extraction and description.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If SIFT method is used to extract keypoints with Difference of Gaussian pyramid, then measurement precision and reliability are improved, but productivity deteriorates due to high computational cost
Solution Approach 1:
The patent changes the mathematical parameters by replacing the computationally expensive Difference of Gaussian filters with Determinant of Hessian filters. This parameter substitution maintains the scale-space detection capability while significantly reducing computational complexity, thereby improving extraction speed without sacrificing detection precision
Solution Approach 2:
The patent substitutes the mechanical filtering process of SIFT (multiple Gaussian convolutions) with a mathematically equivalent but computationally lighter Hessian determinant calculation. This substitution replaces the multi-step mechanical filtering system with a more efficient mathematical operation that achieves the same scale-space analysis goal
2Productivity
If SURF method is used with Haar-wavelets and integral image, then productivity is improved through faster extraction, but measurement precision deteriorates compared to SIFT
Solution Approach 1:
The patent changes the filter parameters from Haar-wavelets (used in SURF) to Determinant of Hessian filters with Gaussian smoothing. This parameter change retains the computational efficiency of direct filter application while improving detection precision by using Gaussian-based scale-space theory, which is more robust to scale variations
3Measurement precision
If complex descriptors like DAISY are implemented with full computation, then measurement precision is improved, but productivity deteriorates due to computational demand
Solution Approach 1:
The patent performs preliminary computation by calculating the Determinant of Hessian and its derivatives during the keypoint detection phase. These pre-computed values are then reused for descriptor calculation, eliminating redundant computations and significantly speeding up the descriptor generation process while maintaining high discrimination ability
Solution Approach 2:
The patent merges the keypoint detection process with the descriptor computation process. By computing the Hessian determinant and its derivatives once during detection and reusing them for descriptor calculation, the patent combines two separate computational stages into a unified process, reducing overall computational demand while maintaining descriptor quality
Data Source
AI summary
Method, system and computer program for efficiently extracting and describing scale-space interest points. It is designed towards low overall computational complexity. On one hand, the data acquired during extraction in the description phase is intensively re-used. On the other hand, an algorithmic optimization of the description that dramatically speeds up the process, is proposed.First, the image is filtered with triangle kernel at different scales. The triangle filtered images are reused for extraction of the keypoints dominant orientation and the computation of the DAISY-like descriptor.