Article Preview
TopIntroduction
Diagnosing an illness is an essential step toward treating an illness. There are a number of screening methods and medical imaging is one of those. It is a technique for creating a conceptual image of the human body's ideal component including muscles, bones, and brain. To do this several multi-modal medical imaging technologies have been implemented. Image fusion is the principal solution among them. Collecting and fusing necessary information from different images into a composite image, without distortion and detail loss. Image fusion has several benefits, including less confusion, enhanced precision, image sharpening, improvement of characteristics, better recognition, early stage detection of illness, simple to read, low expense, minimized data transfer and color fused image. Image fusion may be used in medical care, space imagery, identification and classification of artifacts, and military.
For image fusion, different techniques and different classifications have been proposed in literature. The important techniques which are depends on image spatial domain (Ghassemian, 2016), techniques using any DWT or DCT transform (Jin et al., 2017), technique with contrast, morphological and ratio pyramids (Petrovic & Xydeas, 2004), technique with Laplacian pyramid (Yang et al., 2009), techniques depends related to image fusion in pixel level (Naidu & Rao, 2008), techniques based on image feature (Naidu & Rao, 2008), techniques based on decision level (Al-Tayyan et al., 2017). The fusion method for the spatial domain image interacts specifically with the pixels of the reference image. This approach can be evaluated in four ways: principal component analysis (PCA) (Wan et al., 2013), Intensity hue saturation (IHS), clear average, and clear maxima. They may have some drawbacks while they are really useful techniques. An easy maximum creates images that are strongly oriented but it induces blurring (Xu, 2014). Blurring may influence spatial contrast. Simple averaging is not capable of providing direct images of the target but is the easiest tool for image fusion. According to the principal variable research method, spectral loss exists. Saturation of saturation hues is only appropriate for color image fusion, so it is applied for medical imaging applications (Haddadpour et al., 2017).
In case of techniques depends on transform techniques, the image is transformed from spatial to frequency domain and the resultant frequency coefficients are fused to get fused image of two images. To get back spatial fused image apply inverse transform. In these techniques, images are converted to multi-scale or multi-resolution representation before fusion. These techniques covert images by using discrete time wavelet transform (DT-DWT), Complex WT (CWT), Curvelets (Choi et al., 2005), additive wavelets, non-sub sampled Contourlet transform (NSCT) which is used for generates edges, complex contours and textures (Majhi, 2018). Among all, some techniques in transform domain are very better than spatial techniques in few measuring aspects. The very important, discrete wavelet transform can able to handle curved edges which leads fail in fusing brain images in effective way. These techniques have very bad shift sensitivity, directionality, less spatial resolution and destruction of phase information and because of down-sampling effect there may be chance of Pseudo-Gibbs effect. Where as in Curvelet transform, which are good in handling curved edges and capturing curvilinear. So it is best suitable for brain image fusion at the cost of bearing expansive process and consumes more time in execution. In similar, DT-CWT techniques overcome all the drawbacks occurred with spatial and transforms domain with considerable drawbacks. DT-CWT techniques are good in perfect identification of edges, obey property of shift invariance, and take less time in exaction and better directionality properties (El-Hoseny et al., 2018).