Multiple single pixel dim target detection in infrared image sequence

Abstract

It is difficult to detect and classify multiple point targets in the presence of clutter and noise, in an air borne IR image sequence. Most existing approaches assume a target of several pixels size or of Gaussian shape with a variance of 1.5 pixels. We propose a wavelet based detection technique for multiple single pixel dim targets in an IR image sequence in the presence of clutter and noise. Temporal multiscale decomposition is used to detect temporal change. This temporal change map is postprocessed to reduce the probability of false alarm and misdetection.