COLOR IMAGE IDENTIFICATION BASED ON 2-D POWER SPECTRUM BASED ON NEURAL NETWORK

Abstract

Abstract:Image identification plays a great role in industrial, remote sensing, and militaryapplications. It is concerned with the generation of a signature to the image.This work proposes a dynamic program (use Neural Network) to identify the color imagedepending on the distribution of the monochrome colors (red, green, and blue) in the sameimage to make image signature accordingly, which is represented by a values named powerspectrum. The first step is to analyze the three-band monochrome image (color image) toRed, Green and Blue image, then deal with each image as a grey scale one which isrepresented as a 2-D matrix. The second step is to make Fourier Transform to each greyscale image in order to extract the implicit information in that image. The calculations of 2-D Power Spectrum for each image have been done to construct the final feature vector foreach one. Finally, in the third step, and in order to handle problems of large inputdimensions, a multilayer perceptron Neural Network has been used with two hidden layers.The input of the Neural Network structure is the final feature vectors which are obtainedfrom the previous step. All programs are written using MATLAP VER. 6.5 programminglanguage.