Crate sift_features

This crate contains an implemenation of the SIFT image descriptor. It aims to be compatible with the implementation found in OpenCV’s feature2d module and you should be able to match features extracted with OpenCV and this crate.

Useful resources:

• [1]: Lowe 1999
• [2]: Lowe 2004
• [3]: Mikolajczyk 2004
• [4]: Rey-Otero 2014

The code tries to follow [4] (Anatomy of the SIFT Method) in particular. It deviates in a few places to be compatible with the SIFT implementation OpenCV, namely how histograms are smoothed, angle computations and some details in how the final descriptor vector is calculated.

Structs

ImageprocProcessing

Uses imageproc implementations of gaussian blur and resizing.

KeyPoint

SiftResult

Traits

Processing

Basic image operations used by SIFT. For testing or benchmarking, it’s useful to use exactly the same blur and interpolation procedures to obtain identical results and performance.

Functions

sift

Extract SIFT features using default blur and resize implementations.

sift_with_processing

Extract SIFT features using provided blur and resize implementations.