Study On Non-uniformity Correction Algorithm Of Infrared Imaging And Its FPGA Implementation

As the most important part in a staring infrared imaging system, Infrared Focal Plane Array(IRFPA) has gotten the fastest development in all kind of infrared detectors for the past few years. Because there is non-uniformity between all elements in an IRFPA, it produces the Fixed Pattern Noise(FPN) in infrared image sequences. The non-uniformity is usually caused by the manufacturing imperfection of sensor, the dark-current noise of circuit and the change of environment, and it not only severely degrades the imaging quality of a system, but also has an effect on the further use of infrared images. Therefore, it is necessary to improve the quality of infrared images, and the non-uniformity correction(NUC) technique based on digital image processing is the most effective method. The purpose of this paper is to study a scene-based adaptive non-uniformity correction algorithm and its hardware implementation. Finally, a real-time correction system is designed.By studying on the neural network based non-uniformity correction algorithm, this paper analyzes its existing problems and makes an improvement on it. On the one hand, the guided filter is used to estimate the real value of infrared images, which can be regarded as an edge-preserving smoothing operator. It improves the correction ability of the algorithm and prevents the generation of ghosting artifacts at the same time. On the other hand, a projection-based method is used to detect the movement of scene, and the correction parameters are updated when there is a significant movement between adjacent images. It effectively prevents the fuzziness of the stationary scene caused by invalid updating of parameters. For above improved algorithm, different image sequences are used to simulate the function, and its correction ability is analyzed by calculating the indexes of image roughness, mean square error and signal-to-noise ratio. It turns out that the proposed algorithm has obvious advantages compared with the traditional neural network algorithm.In addition, an FPGA based hardware system is designed in this paper, and the improved algorithm is implemented by using the FPGA’s characteristic of programmable and parallel computation. A real-time adaptive non-uniformity correction system is designed in this paper, and it processes a 256×256-pixel infrared video signal at 25 frames per second. First, the hardware system design is described followed a top-down design methodology,including the top design of FPGA, the partition of function modules, the data flow between different modules and the design details of each module. Next, the function of our system is validated by simulation and test, and the performance of our system is analyzed in two aspects consist of FPGA resource usage and computation speed. Finally, it turns out that the system presented in this paper can correct the non-uniformity of infrared images automatically in real time, and effectively prevent the phenomenon of ghosting artifacts and fuzziness.