A Comparative Review and Optimization of Surf, Brisk and Freak Feature Descriptors

The concept of feature detection is a method to compute abstraction of image information at every point of an image and making local decision at that particular point that there is a feature in an image or not under image processing and computer vision. An interesting part of an image can be called as feature. In this paper I have done the comparison between three keypoint descriptors and have proposed a new combined approach to detect the keypoints present in an image. One of the descriptor is well known as SURF descriptor and other two are new in the fields which are called BRISK and FREAK. In this paper comparison is done between the three descriptor in terms of the Number of Features detected and Time Taken by each of the descriptors to do so. Thus by dividing the time by the number of feature we can calculate the average time taken for a single feature lesser the number greater is Descriptor in terms of speed and lesser computation power. In the end I have made a flow chart of new proposed methodology which I will work to implement that proposed methodology and compare them with the other descriptor in my future work.