Studies on the performance analysis of enhanced image fusion techniques

Abstract

Image fusion is defined as the process of combining relevant information from two or more images captured using different sensors, in order to produce a single output image with an appealing visual perception. It is considered to be one of the most powerful tools revolutionizing the field of image processing in diverse areas like medicine, astronomy, defense and so on. For e.g., fusion of an IR image and visible image finds its application in military field. Fusion of Computed Tomography (CT) and Magnetic Resonance (MR) images provides better information in medical diagnosis. It is really a challenging task to carry out the fusion of two images in an intelligible manner to form a composite image. The key issue is that, the salient information present in the two input images is to be preserved in the composite image formed. Many approaches to image fusion have been attempted in literature, to achieve this goal. There are mainly two groups of fusion methods: spatial domain approach and transform domain approach. Spatial domain techniques suffer from the serious drawback of spatial distortion. This drawback is overcome by using transform domain methods, of which the wavelet based techniques are found to be quite popular. The core of the thesis work is focused on image fusion using Dual Tree Complex Wavelet Packet Transform (DTCWPT). Pixel based image fusion of two images using DTCWPT is carried out to get the final fused output and compared with that of DTCWT approach. To overcome the shortcomings encountered in pixel based fusion, region based fusion using DTCWPT is also attempted. The results obtained are compared with that of the pixel basedtechnique. Further work is concentrated on region based and pixel based colour image fusion using DTCWPT. The next phase of work is concentrated on watermarking the fused image in telemedicine applications for ensuring security measures in information transfer.