Multi-Exposure Image Fusion Based on Illumination Estimation

Enhancing Image Details with Multi-Exposure Fusion

Abstract

High dynamic range (HDR) photos with varying exposures. Pixels in the original picturesequence are analyzed to determine how well exposed they are using illumination estimate filtering.Following these baseline estimates, membership functions are used to give certain pixels in the picturesequence more or lesser importance. Imaging a diatom species with complex silica-based cell walls andmulti-scale patterns with a conventional microscope is insufficient due to the limited dynamic range of thedata collected. Microscopy methods that include taking many photos with varying exposures in order toglean features from the diatom are used. A new breakthrough indicates that image fusion overcomes thelimits of typical digital cameras to capture details from high dynamic range scene or specimen takenutilizing microscopic imaging methods. Visual evaluation and numerical metrics of contrast, brightness,and saturation are used to assess the effectiveness of the suggested strategy. The results demonstratethe method's effectiveness in boosting details without altering the color balance or adding saturationartifacts, and also highlight the value of fusion approaches for picture improvement purposes.Furthermore, we develop an encoder that combines a CNN module with a transformer module to makeup for the shortcoming in building long-range dependencies in CNN-based architectures. Thiscombination allows the network to concentrate on both local and global information. On the most recentiteration of the publicly available multi-exposure picture fusion benchmark dataset, we found that ourtechnique outperformed its rival conventional and deep learning-based counterparts in both subjectiveand objective assessments.