COA: A novel contrast fusion algorithm for GM-APD range image and infrared image under low-photon condition

Most existing image fusion methods focus on infrared images and visible images, which underperform under complex conditions. However, they cannot effectively distinguish similar targets at near distances and it limits the target recognition, especially when the image is missing badly under low photon conditions. To solve this problem, we propose a contrast optimization algorithm (COA) for fusing the Geiger-mode avalanche photodiode (GM-APD) image and infrared image that could emphasize and distinguish similar targets in low photon conditions. The COA method includes two main parts. First, using the processed infrared image as a reference to recover the GM-APD image by using DLatLRR system methods and reference recovery method. Then, we proposed a new loss function consist of contrast fidelity term and regularization term which focus on the contrasts of the infrared image and GM-APD image, and introducing the split Bregman to solve the optimization equation. Results show that the value of contrasts between targets (TC) of the proposed method is twenty times more than the infrared image and higher than that of the other 9 fusion algorithms when the mean photon per pixel (PPP) is 1/500. Meanwhile, the proposed method can also be used for combined with the existing fusion methods to fuse the infrared image and GM-APD image, which makes accurate target identification under low photon conditions.