Development of New Image Fusion Techniques

Abstract

The image fusion process aims to combine the crucial information from multiple images obtained from different resources. Numerous image fusion techniques are proposed to achieve better fusion accuracy however they generally yield less informative edge details, introduce artifacts and have blurred details in multi focus images. More- over, in medical images, resultant fused images have missing fine details of different tissues, improper fusion of different tissues and boundaries are also not clearly demarcated. This work presents, different multi-focus and multi-model techniques to minimize the above-mentioned issues/limitation. Cross bilateral filter which provides better perceptual quality is combined with non-sub sampled contourlet transform to maximize the useful information present in multiple scales and directions. Guided filter whose computing time is independent of filter size and well preserves edges is combined with discrete wavelet transform for more in formative edge details. Alpha map along with top hat transform extracts common properties of different images. Saliency map shows more meaningful representation of image therefore it is used along with cross bilateral filter to fuse multimodal brain image modalities which help to combine the facts found in multiple medical images that have exceptional distinctive modalities. Energy of Laplacian and discrete wavelet transform based fusion of multimodal brain images helps to properly demarcate boundaries of different tissues. Shift invariant discrete wavelet transform and sparse fusion is used for more informative anatomy details of medical images which helps radiologists in evaluating different modality images. Proposed multifocus image fusion schemes maintain different image details like reduce edge blurring, introduce less artifacts and preserve sharp edge details. Moreover, suggested multimodal image fusion schemes for medical images preserve the fine details of different tissues, proper fusion of different tissues and clearly demarcate the boundaries to obtain high quality fused image. Higher quantitative and qualitative out- comes are observed in the proposed frame works for fusion schemes as compared to other existing schemes. In general our research work presents multi-focus and multi-model image fusion problems in two main domains i.e. photography applications(to overcome the problem of depth of field of cameras) and biomedical imaging. The proposed novel and effective algorithms have been tested on real world data to demonstrate their performance improvement.