Kernel Principal Component Analysis feature detection and classification for underwater images

Abstract

This paper examines the Kernel Principal Component Analysis (KPCA) feature detection and classification for underwater images. In Underwater images the numbers of distortion occurred are blurring of image, illumination of light and rotation of angle, noise etc. Features are normally extracted by the method called SIFT (Scale Invariant Feature Transform for underwater images). It is used for extracting distinctive invariant features from images that can be invariant to image scale and rotation. It is used in image mosaic, recognition, retrieval and etc. where PCA-SIFT (Principal Component Analysis-Scale Invariant Feature Transform) is also used for dimension reduction and feature detection for underwater images. In this paper we propose a method by combining KPCA and SIFT together called KPCA-SIFT feature detection for underwater images. It is well suited for blur, illumination change and rotation of the input image. When apply PCA to the normalized gradient patch it reduces the dimension of the feature extracted. The parameters used for evaluation are Precision Vs Recall curve and the parameter like elapsed time, sigma value and threshold value. PNN classification is used for underwater images. The proposed method gives desirable results with respect to parameters like Roc curve, Precision Vs Recall, Elapsed time and sigma value. It is more robust and distinctive image deformation and more compact. with increased accuracy and faster matching than the PCA SIFT and SIFT algorithms.