An adapting object detection of infrared image based on optimal hybrid threshold surface

Due to the low signal-to-noise ratio and the relatively small objects of infrared image, we propose a novel improved object detection algorithm. In our algorithm three customer variables such as Gaussian background model deviation (GBMD), relative radiation intensity difference (RRID) and region correlation (RC) have been defined to describe information of the target local region texture characteristics, simultaneously local regional gray distribution and adjacent regionspsila correlation information can be used effectively. At last we will get a hybrid threshold surface, with its help the image can be automatically divided into two classes (background and target). Experiments indicate that our algorithm is good at using image information and its detection efficiency and accuracy have been improved.