Infrared Dim Target Detection Under Sky-based Background

Infrared dim target detection under the sky-based background is major studied inthis paper. First of all, in order to improve Infrared (IR) image quality, blind-pixeldetection and compensation for IRFPA is studied in-depth. Based on the generationmechanism and response characteristics for the blind-pixel, a fast and effectiveblind-pixel detection algorithm, which is double threshold iterative algorithm, ispresented. After that, the response-value of blind-pixel of total line is replaced by theneighbor line pixel, and the isolated blind-pixel is compensated by the interpolationusing the response-value of the four neighbor pixels. Then, based on single-frame image,two types of algorithm on IR dim target detection are focused. One is two stepsprocessing algorithm. In the first step, a collection of potential targets is extracted. Thenin the second step, the real target is extracted by judging and comparing for the potentialtargets. The other detection algorithm is the adaptive anisotropic filtering method. Inthis algorithm, the filter is based on a modified partial differential equation. Thebackground is pre-estimated by selecting appropriate filter coefficients, and then thetarget signal is separated from the filtered image by adopting a proper threshold. Finally,because the dim target and the random noise have some similar features in thesingle-frame image, in order to solve the difficult problem in separating dim target fromrandom noise, the continuous property of the motion-position and the relativity ofgray-scale-value of the target in the sequence image are utilized, and a new dim-targetdetection algorithm with the image sequence is presented. In the presented algorithm,the max-median filtering algorithm and the kalman-filtering algorithm are combined.The experimental results demonstrate that the presented method is effective for diminfrared target detection under the sky-based background.