A Remote Sensing Image Fusion Method Combining Low-Level Visual Features and Parameter-Adaptive Dual-Channel Pulse-Coupled Neural Network

Remote sensing image fusion can effectively solve the inherent contradiction between spatial resolution and spectral of imaging systems. At present, methods remote images based on multi-scale transform usually set rules according to local feature information pulse-coupled neural network (PCNN), but there are problems such as single feature, rule cannot extract information, PCNN parameter setting is complex, correlation poor. To this end, a method that combines low-level visual features parameter-adaptive dual-channel (PADCPCNN) in non-subsampled shearlet (NSST) domain proposed paper. In low-frequency sub-band process, constructed by combining three features, phase congruency, abrupt measure, energy enhance extraction ability information. process high-frequency fusion, structure parameters (DCPCNN) optimized, including: (1) morphological gradient used an external stimulus DCPCNN; (2) implement representation difference box-counting, Otsu threshold, intensity complexity setting. Five sets data different satellite platforms ground objects selected for experiments. The compared with 16 other evaluated from qualitative quantitative aspects. experimental results show that, average value sub-optimal five data, optimized 0.006, 0.009, 0.035, 0.037, 0.042, 0.020, respectively, seven evaluation indexes entropy, mutual gradient, frequency, distortion, ERGAS, fidelity, indicating has best effect.