The Enhancement Of Nighttime Video Image By Fusing Different Sensors

With the development of science and technology, the quality of the video have been demanded increasingly. The crime rate in 2014 is 2.5 times as much as that in 2013 at China,which 52.3 percent growth rate is occurring at night, mainly because the lights and quality are lower, the shot video which captured in the night almost all black, and we can't recognize the target and background from video, thus there exist challenges for police's investigation, and it's hard quickly control crime. Border defence always existed the phenomenon of illegal immigration, the smuggling in 2014 is 4.75 times in 2013, which 67.3percent growth rate is happening in the evening, in the same principle, illegal immigration is easy to hide and smooth sneak at night. The above phenomenon occurs because the image was captured by single sensor has a little date and information, so it's difficult to enhance it.But it's essay to enhance the image which was captured by multiple sensors because it has enrich data and information. Thus we proposed nighttime enhancement video system based on the fusion of infrared and visible imaging sensors.In this paper, according to the characteristics of infrared images and visible images,we proposed a nighttime enhancement video image system based on infrared imaging sensor and visible imaging sensor were fixed and capturing images at the same time. The algorithm can be divided into two stages. The first step is to calculate the transformational matrix between the infrared video and visible video by the improved sift algorithm. The transformational matrix was used to register infrared video and visible video, and it has high precision and speed to register features. The second step is to fuse infrared video and visible video by ? weighted algorithm based on transformational matrix. Since infrared and visible images have different resolutions, so we should interpolate them for every frame of the video by the transformational matrix, and then fused interoplated infrared and visible images by? weighted algorithm, resulting the time of fusion of every frame at about 30 ms. Thus it achieved good real-time effect.