AND AFFINE INVARIANT(SAI) DESCRIPTORS FOR MATCHING COLOR

Image matching has a vital role in many of the present day computer vision applications such as tracking, image indexing and retrieval etc.,. The problem in image matching is finding the features and descriptors which are invariant to the transformations that the image is subjected to. The descriptors should be of low dimension for real time applications. Thus they should be chosen such that a tradeoff is achieved between accuracy and speed. This thesis deals with a novel method of matching color images under Scaling and Affine transformations. We derive an equation which shows the invariance property of both Affine and Scaling transformation. Using this property we construct a three dimensional descriptor, which we called Scale and Affine Invariant(SAI) descriptor. We have used the three color channels separately in order to get sufficient equations to solve for the Affine parameters. From this we obtained an invariance property for Affine transformation. We then integrate this wit