Detecting Features via Modified Harris Corners and Matching Them through SIFT

Interpreting images spatially is a daunting task which is achieved by detecting corners and features.The most important task of detecting features is achieved by Harris Corner Algorithm. The algorithmis not robust to different scale of the same image. The algorithm may detect corner but when theimage is zoomed in, the corner may appear as ridges. We use the corners detected from HarrisCorner algorithm and treat these as key points to pass into Scale Invariant Feature Transform (SIFT)algorithm. The SIFT algorithm extracts descriptor vector of dimension 128 X 1 from these cornersand can be used to find similarity between different images. This process is quite robust to noise,intensity, scale and occlusion and is used for matching images from a database of descriptors. Wehave investigated both the algorithms in this paper and made a modified version of Harris Corneralgorithm by performing different kind of thresholding, both of them gave a little different result.