Thermal images regularly endure from small resolution, noise, and limited contrast, hindering their practical applications. Due to these problems, we introduce a present research work on thermal advancement method that leverages multi-modal fusion and sparse representation techniques, coupled with deep learning. Our approach involves fusing thermal images with complementary information from visible light images to improve image quality. We employ a deep CNN to uproot the highest aspect from both modalities, sequentially a sparse representation layer to reconstruct the enhanced thermal image. The sparse representation layer encourages a sparse rendering of the fused features, leading to a more robust and efficient enhancement process. Extensive experiments illustrate the dominance of present work extinct ultramodern techniques in terms of both quantitative and qualitative evaluations.