Comparison of Different Feature Detection Techniques for Image Mosaicing

Image mosaicing is widely used in present computer vision applications. A considerable measure of important information is represented by the feature points in an image. Accurate extraction of these features is an essential part of image mosaicing as it can reduce misalignment errors in the final mosaic. A number of feature detection algorithms have been developed in recent years which can be used for image mosaicing. However, the computational complexity and accuracy of feature matches limits the applicability of these algorithms. In this paper, four widely used feature detection algorithms, Harris, SURF (Speeded-Up Robust Features), FAST (Features from Accelerated Segment) and FREAK (Fast Retina Key point) feature detection algorithms are compared in terms of accuracy and time complexity for mosaicing of images correctly. First, these algorithms have been applied on a single image and then, different set of images are tested for the comparison. It is concluded that the FREAK algorithm is superior to the rest of the feature detection algorithm in terms of accuracy and run time.