Performance Analysis of Feature based Multiple Descriptors with Histogram Equalization for Image Mosaicing

Abstract

Objectives:The main objective of the proposed work is to develop an image mosaicing model for combining the images of different individual images. In other way, the union of two images and to evaluate the performance of the model in terms of the number of run time in seconds and number of key features they use. Methods: In this work, the Histogram Equalization is a processing step required to make the mosaic invariant to intra image and inter image intensity variability. The detailed feature of the enhanced image is extracted using Scale-invariant feature transform (SIFT), Oriented Fast and Rotated Brief (ORB), Binary Robust invariant scalable key points (BRISK) feature descriptors techniques. The features including local and global are matched using K-Nearest Neighbor. Then, homography is performed using RANdom SAmple Consensus (RANSAC) algorithm to compute the camera motion. Finally, the image warping is performed using smoothing filter of size 40x40 to obtain the panorama image. Findings: The model is tested on various datasets using three different feature extractors popularly used in image mosaicing or image stitching algorithms SIFT, ORB and BRISK descriptors. It is observed that the ORB is the best feature extractor among the state-of-the-art feature extractors. The ORB with HE uses a minimum of 500 key features to match the images and generates panoramic images that are invariant to shift and rotation with the minimum run time of 0.0836 seconds. Novelty: The state-of-the-art models developed by the researchers suffer with good number of matching points of the input images to generate the image mosaicing. This issue is addressed in the proposed model using multiple descriptors SIFT, ORB and BRISK techniques. The ORB with HE will provide the minimum key features and enough good matches of features with the record of minimum runtime to obtain the panoramic images.https://www.indjst.org/