Redundant Keypoint Elimination Using Redundancy Index on Basic SIFT Algorithm

Abstract

ABSTRACT
Matching features across different images is a common problem in computer vision. For the images of different scales and rotations, there is the need of scale-invariant feature transform (SIFT). SIFT identifies the distinctive keypoints that are invariant to location, scale, rotation, robust to affine transformations (changes in scale, rotation, shear, and position) and changes in illumination, which are usable for object recognition. SIFT algorithm includes two main stages: feature detection and descriptor extraction. The feature detection is performed in four phases: (i) extracting scale-space extrema, (ii) improving accuracy of localization, (iii) eliminating unstable extrema, and at the end, (iv) allocation of orientation to each created feature. This project aims to identify and extract the keypoints using the SIFT algorithm based on reference image. The number of extracted keypoints is based on the complexity of image locations. Descriptor extracts the keypoints from reference image. Euclidean distances are calculated between each keypoint and all other keypoints in the reference image. Threshold value must be selected to remove redundant keypoints. The smaller distance value belongs to the more redundant keypoint and the larger distance value elimination improves the quality of the method. This project work is simulated in MATLAB 2013a for various images, and compared with the existing SIFT algorithm, optimum keypoint extraction will be performed in future.

Keywords: Euclidean distance, object recognition, SIFT

Cite this Article: N. Manikandaprabu, S. Vijayachitra. Redundant Keypoint Elimination Using Redundancy Index on Basic SIFT Algorithmtection. International Journal of Image Processing and Pattern Recognition. 2019; 5 (1): 1-5p.