Algorithm Design of Infrared Image Three-Dimensional Non-Stationary Noise Based on Time Series Statistics

With the development of high orbit area array starting infrared imaging technology in China, the quality of its image directly affects the user’s use. Therefore, noise detection and supplementation of area array staring infrared images are becoming increasingly important. This paper presented a detection method of the infrared image of three-dimensional non-stationary noise based on Time Series Statistics to solve the effect of the three-dimensional non-stationary noise, which formed because many factors influenced the infrared image of staring Plane Array such as orbit time, the inherent flaw of detectors, and imaging conditions. Meanwhile, this paper also located and analyzed the non-stationary noise, and suggested a method of Local filtering by using the trend extrapolation theory. Based on the evaluation of the test result, the non-stationary noise recognition accuracy of this method could be up to 85% and the non-stationary noise image correlation coefficient before and after removal could be up to 95%, which could provide the satellite with real-time processing and improve the quality of the infrared image. This method can effectively improve the data quality and information product quality of high orbit array infrared remote sensing images, promote application level, and enhance application efficiency.