7th International Conference on ICTCS 2022

Authors - Nalini S Jagtap, Sudeep D Theapde
Abstract - One of the developing facets of technology is the picture fusion process, which, by employing a variety of methods, makes it possible to improve the capability of interpreting image features. Fusion of the image is the performance of integrating data from multiple pictures into a one image. Image fusion is gaining momentum in a wide variety of critically essential image analytics application domains. This work presents an image fusion system that maximizes the details by employing a contrast enhancement approach that is based on histogram equalization and PCA & GF. The goal of the system is to minimize substantial artefacts and improvise the contrast. For the purpose of weight map development, the suggested technique makes use of a CNN-based network. The first step entails histogram equalization, which is then followed by decomposition of the initial picture through the use of principal component analysis (PCA). In the first stage, the principal component analysis reduces the dimensionality of the picture while maintaining all of the vital information contained within it and generates an up-dated weight map. At the result of PCA, a Guided Filter is used in order to pre-serve the edges and further accentuate the features, hence eliminating the blurring artefacts. During the third stage of the process, a CNN network that has already been trained generates a new weight map by identifying crucial aspects of images contained within the input dataset. The output feature map is created by combining the weight maps that were generated by the GF and the CNN. These weight maps are then fused to reference picture to produce the output fused image.