Development of Image Stitching Using Feature Detection and Feature Matching Techniques

Image stitching is a process of merging two or more images of the scene into single image of high resolution which is also termed as panoramic image. Image stitching is used to assimilate information from several images by overlapping view fields to create a panoramic view without loss of any information. Stitching of images can be performed by two types of common approaches such as direct and indirect techniques. Direct techniques involve direct comparison of image pixel intensities that are combining. Indirect techniques are dependent on image features. These techniques incorporate feature detection and feature matching of the images to be stitched. In this proposed work, efficient feature detection and feature matching techniques such as Scale-Invariant Feature Transform (SIFT), Speeded-Up Robust Features (SURF) and Features from Accelerated Segment Test (FAST), Euclidean distance and Random sample consensus (RANSAC) are used for the process of image stitching. Images of a scene are captured with fifteen degree difference. These images are pre-processed if needed and then fed to feature detection and matching process. Using the matched feature points, images are stitched to get a 360 degree scene in a single image as outcome. The FAST technique gives good matching points of images using RANSAC which help in obtaining a better stitched image with integrating all the data of different images involved in stitching. The experiment is carried out for two datasets and average detected feature points and matched points for three algorithms are computed. The SIFT algorithm gives 36 matched points among 319 detected points, SURF gives 55 matched points among 334 detected feature points and FAST gives 102 matched points among 774 detected feature points. Image stitching finds application in various fields such as medical imaging, satellite imagery, automobile industries, image stabilization, document mosaicing and further image stitching can be extended to video stitching.