Research on Multimodal Image Fusion Method Based on Deep Learning

Accurate assessment of medical image quality is of paramount importance as it forms the cornerstone of diagnoses, where even minor discrepancies can lead to misdiagnoses. To achieve this, two key aspects, realism, and comprehensibility, are employed to evaluate the quality of medical images. Factors such as image richness, clarity, and saliency are analyzed to precisely reflect the image's reliability. However, due to variations in data originating from different hospitals, medical professionals, and equipment, internal structures may exhibit significant differences. To address this issue, this study proposes a multimodal data-based medical image fusion model that captures the correlation and complementary information between heterogeneous data. By combining the regional Laplacian pyramid (LLP) with convolutional neural networks, the convolutional network is enhanced. The source images are decomposed using LLP, and an improved deep convolutional neural network generates an optimal weight map to guide the fusion process, resulting in a fused image. Experimental results demonstrate that the proposed method achieves favorable fusion effects in both subjective visual assessments and objective evaluation metrics.