Intelligent Infrared Image Non-uniformity Correction Algorithm And Its FPGA Implementation

The infrared imaging technology has been widely used because of the good development. But the deficiency of manufacturing technology results in the non-uniformity, which degrades image quality. It’s necessary to develop effective non-uniformity correction (NUC) methods and technology. The NUC algorithms have been proposed can be divided into two categories roughly:the first category is based on calibration, which can’t correct the parameters adaptively and hard to meet people’s demands. The second category is based on scenes, which are immature, and there is no reference about how to realize them on hardware.This paper analyzes the reasons cause the non-uniformity of infrared imaging system, and some common NUC methods like calibration methods, neural network correction method and constant constant-statistics method are simulated and analyzed. Finally, an intelligent NUC method based on gray projection registration is proposed. The method calculates the motion vectors by projection values and cross-correlation function, and corrects the gain and offset parameters according to the method of steepest descent. Moreover, the ghosting problem is solved by a ghosting restraining method. The performance presented by software simulation and FPGA hardware correction experiment proves that the new method has quick convergence speed, good correction without ghosting problems, strong adaptability and stability.