Extreme Learning Machine for Effective Medical Image Retrieval

Abstract

This paper presents an adaptive methodology for medical image retrieval using varies rotations. Dominant rotated local binary pattern has been tested in last few years in matching the images and is good choice to check for retrieval of medical images. The dominant rotation direction is considered, when the index of difference in distance of central pixel and neighbouring pixel is maximum. The direction of dominant path is considered in circular path and assigned the weights with the help of the dominant path. Because of the having dominant path direction, this gives faster, easy and efficient results in retrieving process. Histogram of rotational features, gradients, curvature features and various features of the query image and data images have to be compared to conclude a best suited image or images to query image.