Comparison of infrared image and visible image fusion methods

Infrared imaging can provide high-quality images for different conditions and can distinguish targets from their backgrounds based on thermal difference. Visible imaging can provide high-resolution and texture details in accordance with the human visual system. Therefore, it is desirable to fuse advantages of thermal radiation and detailed tissue information. In this thesis, an overview of the main applications of infrared and visible image fusion and all methods used is presented and selected methods are compared. As image fusion methods: DSWT, SR, PCA, Saliency Based, Hybrid (Curvelet and SR) and Infrared Feature Extraction methods have been selected. Then, the quality evaluation metrics that compare the fusion performance of each metric have been discussed in detail. For comparison, 6 methods and 9 evaluation metrics have been used practically by simulating in MATLAB™. According to results of visual quality evaluation criterions: DSWT method gave the lowest result in terms of brightness. The SR method has caused distortions for fused image. The PCA method gave the highest result in terms of brightness. Saliency-based method gave the best visual results for 3 source image sets. Since hybrid method over-transfers infrared image information to the fused image, visual detail information is distorted. Since infrared feature extraction method focused only on infrared feature, it gave an average result without concentrating on other visual information. According to results of objective quality evaluation metrics: DSWT and SR were superior in 3 of 12 pairs source image sets. PCA was superior in other 6 of 12 pairs source image sets.