Non-uniformity correction for infrared focal plane array with image based on neural network algorithm

摘要

Non-uniformity response of detectors based on infrared focal plane array (IRFPA) result in fixed pattern noise (FPN) due to detector materials' non-uniformity and fabrication technology. Once fixed pattern noise added to the infrared image, focal plane image quality will have a serious impact. So non-uniformity correction (NUC) is a key technology in IRFPA application. This paper briefly introduces the traditional neural network algorithm and puts forward an improved algorithm for the neural network algorithm for NUC of infrared focal plane arrays. The main improvement is focused on the estimation method of desired image. The algorithm is used to analyze the image array, correcting data on the array both in space and in time. The correction image in the text is from the infrared data sequence which is more successful of three frames of data obtained. It was found that the estimated image corrected by new algorithm is closer to real image than the estimated image corrected by other algorithm. Moreover, we simulated the new proposed algorithm using Matlab. The results showed that the method of spatial and temporal co-correction of the images is more realistic than the original image.