Infrared and visible image fusion based on shiftable complex directional pyramid transform and SUSAN edge detector

We propose an algorithm for fusing infrared and visible images, which is based on a shiftable complex directional pyramid transform and a SUSAN edge detector. When fusing the low-pass sub-band coefficients, we employ a weighted-averaging rule based on local normalized energy. When fusing the directional high-pass sub-bands coefficients, the SUSAN edge detector is utilized to generate a decision map and guide the fusion process. In order to reduce the computational complexity, we calculate the SUSAN edge response only once directly for each source image. We evaluate our fusion algorithm on a standard TNO Image Fusion Dataset, using comparisons with a number of traditional fusion algorithms. The experimental results testify that our algorithm is efficient and feasible. Moreover, it is superior to the traditional algorithms from the viewpoints of subjective evaluation and objective fusion metrics.