Image fusion using online convolutional sparse coding

As signal enhancement technique, image fusion alleviates limitation single sensor in terms to information presentation and enhance visual quality. Extracting affluent features to accurately represent image is crucial for fusion. However, filters via convolutional sparse coding (CSC) have disadvantages of heavy computation cost and low representation. Superior signal representation and low spatial complexity of online convolutional sparse coding are exploited to image fusion to compensate for shortcomings of CSC. The detail and low-frequency components of source images are firstly decomposed using two-layer decomposition. Then each layers use rules to obtain fused components. Finally, fused image can be reconstructed by both high-frequency and low-frequency layers. To verify performance of proposed method, 9 infrared-visible fusion methods and 5 medical fusion methods are used as comparison experiments. The quantitative ( Q ABF , Q E , Q M and Q P ) assessments confirm superiority of method. In addition, qualitative results exhibit powerful information preservation and better visualization.