Strip Non-uniformity Correction In Infrared Focal Plane Array

In recent years, with the more and more mature of infrared thermal imaging technology, it has been widely used in many military and civil fields. However, because of the limitation of manufacturing process and material to make in the infrared image usually contains obvious stripe non-uniformity, the image has different shades of stripe noise when it was enhanced, it led directly to the effect of the original infrared images is not ideal. Therefore, eliminating stripe non-uniformity in infrared image and keeping the details of images is an important step of the infrared image processing.This paper introduces the generating mechanism of stripe non-uniformity. Because the stripe non-uniformity is caused by pixels in the same column share the same readout circuit, the strip non-uniformity of pixels in the same column have high correlation. But the influence of the infrared image preprocessing algorithm would lead to the same column stripe non-uniformity exist a little differences, which is the basis of the algorithm in this paper. The shortage of the existing algorithms are not fully meet the requirements of the engineering application, so in this paper presents a correction algorithm based on partial correlation which base on the local correlation of stripe non-uniformity and theoretical basis of the existing correction algorithm. This algorithm will be from the following three parts. First of all, the calibration model of many current correction algorithms adjust the same gain coefficient and offset of pixels in the same column pixels, on the contrast the calibration model of this paper is adjust the gain coefficient and offset of each pixels. Secondly, it use the LMS to remove strip non- uniformity and select the output value of gauss filter to act as predictive value, then to inherit the gain coefficient and offset of previous line and use the steepest descent method to recursion iteration update correction parameters, which aims to making the error between the corrected pixel value and the forecast value minimum. Finally, in the process of recursive iteration add a step about the image edge detection, the correction parameters of previous line of the target pixel in the same column can approximate equal the correction parameters of the target pixel, when the target pixel is located in the image edge.In this paper we select several scenarios of infrared images which contain stripe non-uniformity and experiment results show that the proposed correction algorithm can effectively eliminate stripe non-uniformity, and it keeps the details of the image well. In addition, the correction algorithm proposed in this paper can use single image to estimate correction parameters, and it has linear time complexity and high real-time performance.