Isolation and comparison of features points on the images of object

Isolation points of interest (POI) and determination of their compliance are important components of many applications of computer vision. For example, in the algorithms of automatic alignment and alignment of images, construction («matching») composite mosaics and panoramas, the reconstruction of three-dimensional images, etc. By selecting and implementing a specific algorithm selection and compare features directly affects the speed and accuracy of the reconstruction algorithms. In this article analyzes existing methods of selection and comparing POI, and also is offered the new algorithm, allowing to increase an amount of true correspondences.