Novel Fusion Approaches for Enhanced Image Processing with Smoothed Images

Abstract

Image fusion is integrating multiple images from many sources and changing them into a single image with clearer and more accurate information. Image fusion techniques have been proposed to enhance distorted input images using a smooth filter to improve the clarity of distorted images. This work fused images resulting from smooth filters (half left and half right) with size windows of (3×3), (5×5), (7×7), (9×9), and (11×11) pixels. The image resulting from the smooth filter towards the right was combined with the image from the smooth filter towards the left using traditional techniques such as addition, multiplication, and new suggested techniques, namely absolute real standard deviation, binary standard deviation, real covariance, and binary covariance. The data examined by quality assessment methods with reference depend on Mutual Information, Correlation Coefficient, Structural Similarity Index metric, Structural Content, Normalized Cross Correlation, and without references like Blind Reference less Image Spatial Quality Evaluator, Naturalness Image Quality Evaluator, Perception-based Image Quality Evaluator, and Entropy. Lena's image shows a different behavior than the cameraman and the personal images because Lena's image has more details, resolution, and sharper contrasts. The best combination method was binary standard division.