A Brief Study of Different Feature Detector and Descriptor

Abstract: Robust, accurate and automatic registration of image is complex task in various applications. To achieve image alignment/registration, required steps are: detection of features, Feature matching, source of transformation function. Extraction of different consistent features from image is very useful and that may be used to gain reliable matching between various views of a scene or object. In addition of deciding where and to what range a feature occur in an image, there is a distinctive body of research to decide how to represent the neighbourhood of pixels near a localized region, called feature descriptor. The easiest technique is scaled for the size of the region, use of pixel intensity values, or representation of eigen space. This survey contains various descriptors like SIFT( Scale-invariant feature transformation), SURF( Speeded up robust features), HOG( histogram of gradient), MSER.

Keywords: Key point, SIFT, Feature description, Feature extraction.