Abstract

Image processing techniques have witnessed increased usage in various real world applications. For any image processing technique, such as image segmentation, restoration, edge detection, stereo matching etc., to be applied successfully, the image under consideration must contain all of the scene objects in focus. Usually, due to inadequate depth of field of optical lenses, especially with larger focal length, it becomes impossible to obtain an image in which all of the objects are in focus. Image fusion deals with creating an image by combining portions from other images to obtain an image in which all of the objects are in focus. In this paper, a novel feature-level multi-focus image fusion technique has been proposed which fuses multi-focus images using classification. Ten pairs of multi-focus images are first divided into blocks. The optimal block size for every image is found adaptively. The block feature vectors are fed to feed forward neural network. The trained neural network is then used to fuse any pair of multi-focus images. The results of extensive experimentation performed are presented to highlight the efficiency and usefulness of the proposed technique.

Key words: Multi-focus image fusion, feed forward neural network, feature classification, genetic algorithm.

Abbreviation

DWT, Discrete wavelet; GA, genetic algorithm; EOG, Energy of gradient;SF, spatial frequency; SD, standard deviation; MI, mutual information; PSNR, peak signal to noise ration; RMSE, root mean square error.