21-08-2014, 10:29 AM
With the availability of multi-sensor data in many fields, such as remote sensing, medical imaging or machine vision, image fusion has emerged as a new and promising research area. Image fusion refers to the techniques that integrate complementary information from multiple images or sensors such that the new images are more suitable for the purpose of human visual perception and other computing tasks. Different fusion algorithms have been proposed and implemented to obtain better fused image from the images taken at different conditions and situations.
In this report we focus on the image fusion using the Discrete Wavelet Transform (DWT) as DWT offers more precise way of image analysis than the other fusion techniques. It decomposes images into the low and high frequency sub bands and implements the fusion algorithm on these sub bands. At first we consider pixel by pixel fusion for the Low frequency band and consider multi focus images one of which is focused well at the right side and the other is focused well at the left side. The fusion algorithm we implement here is based on the maximum selection scheme for the details and for the approximate we consider the mean of the two values. This fusion method enhances the image quality. However all the features of an image are not always contained in a single pixel. So we consider area based fusion as a part of our second algorithm where the window size of 5×5 matrix have been defined and the maximum selection scheme is implemented on the same calculating the variance of each block. The reconstructed image contains the pixel value corresponding to the centre pixel having the largest value of variance. Also a binary decision map has been defined to view the consistency of the fused images. We implement the total algorithm in the MATLAB to obtain an enhanced image from two differently focused images. In one case the multi-focused image is considered and in the other the multi sensor image is considered for fusion.