Multifocus Image Fusion using TE-CSR Technique

Image fusion contains data from different heterogeneous sources to obtain more accurate information about an inherent scene. There are various methods are available to increase the quality of an image for better visualization of the input image. The last few years, the familiar technique is used which provides efficient image fusion at the output called Sparse Representation (SR) with number if effective algorithms being proposed. But there are two drawbacks of SR methods, they are high noise sensitivity and the preservation of image details for further processing. To get the better of these drawbacks, the Convolutional Sparse Representation with Gradient Penalties (CSRGP) technique is adopted, which is a signal decomposition model used to perform image fusion of the deviation components of the processing frame. The energy components of an image are also combined using the technique called averaging. To perform the reconstruction of an image after it is being fused, the inverse Taylor expansion is adopted. There is a gap between the image descriptions, if the existing decomposition algorithms are used. It can be suppressed by the proposed method by introducing the new method which can preserve the image details. The sparse patch coding plays an important role to limit the ability to preserve image details along with the SR based approaches. The efficiency and performance measure of the proposed method is demonstrated with the experimental results.