A Novel Medical Image Fusion Approach Based on Interval-Gradient Decomposition and Sparse Saliency

This is the authors' abstract. We don't add key points for this paper.

The medical image fusion method can provide more comprehensive and accurate diagnostic information by fusing multimodal medical images. This technology can help doctors understand patients’ conditions better to develop more accurate treatment plans. It can also be applied to knowledge graphs and data processing. In recent years, a growing body of literature has proposed medical image fusion algorithms; however, these methods still suffer from notable limitations, such as information loss: certain fusion algorithms may cause some image information to be lost or blurred, especially with different modalities; contrast imbalance: some algorithms cannot balance the contrast between images correctly, resulting in some areas being too bright or dim; and artifact generation: some algorithms may produce artifacts during image fusion, which may interfere with the doctor’s interpretation and diagnosis. This study introduces an innovative methodology that leverages interval-gradient decomposition and sparse saliency detection to overcome these drawbacks. By utilizing an interval gradient, the proposed method decomposes original images into structural and texture components, then extracts saliency images of the original image via sparse saliency and finally fuses structural, texture and saliency images via various methods. Finally, the three fusion images are combined to obtain the final fusion image. Comparative experiments confirm that the proposed method yields fusion results with superior detail retention and more preserved salient features compared to baseline approaches.