IMAGE FUSION QUALITY ASSESSMENT USING SHANNON ENTROPY AND HARTLEY USEFUL INFORMATION COEFFICIENT

Abstract

The article examines methods for improving the quality of images obtained from different sources of information through multi-modal integration. By combining information from various modalities, features that cannot be accurately interpreted when analyzed separately can be utilized. To support the relevance of this topic, recent research in this area is discussed. The goal of the work is to enhance the information content of images resulting from merging data from diverse sources and create high-quality images suitable for accurate machine learning applications. To achieve this objective, the authors address several tasks. They develop an approach for measuring image quality and design algorithms to evaluate the quality of fused results based on multi-modal information. These algorithms are implemented within a software framework for validating the proposed approach. Evaluation experiments are conducted based on the presented calculations of the information content of images and the effect of noise and blur on the entropy of the combined image. The results of the experimental studies on data sets from open sources have shown that the proposed method allows determining the best way to merge images with maximum information content. The use of Shannon entropy makes it possible to calculate the amount of information transmitted in images, and the Hartley coefficient of useful information helps estimate the amount of noise in an image. Additionally, the article compares the results at different levels of noise and degrees of blur in images, demonstrating the effectiveness of different algorithms for evaluating image quality. To illustrate the proposed approach, we analyze images obtained by combining data from two devices: an infrared camera and a video camera capturing images in the visible range