Dim target detection system based on local texture analysis and hardware implementation with DSP

In infrared small target images, the target has different texture characteristic and frequency characteristic from background. Using multi-scale wavelet theory, small target is separated from background. A local texture analysis method which is very fit for dim target detection in low SNR (signal to noise rate) is presented to realize the detection. In order to satisfy the real-time realization request of dim target detection, a dim target detection system is presented, and it is based on a kind of high performance digital multimedia DSP chip produced by TI Company. The program's processing speed and pipelining efficiency are improved by optimizing the software program, so the system has a very good software and hardware system structure. Experiments on real infrared image sequences show that the system can detect real time steadily dim target with 1-3 pixels in clutter background.