Infrared Dim Targets Detection In Complex Background

To explore the algorithm of dim targets detection in complex background is helpful for the battle now and latter. And it is beneficial for the further development of precision guided munition. This paper mainly discusses the detection of dim targets in the images with complex background.Image signals can be classified to Gaussian images and Non-Gaussian ones. They can also be classified to stationary images and nonstationary ones. Comparatively, targets in the images which are both Gaussian and stationary are easy to be detected. There has been perfect algorithm for this. Here we deeply study the targets detection in non-Gaussian and nonstationary images. Least-squares estimation is used to acquire the optimal correlation length of the neighborhood of every pixel. A measure in relation to correlation length is taken to assurance the correlation matrices to be reversible. We design a new algorithm which can be used for the infrared dim targets detection in the non-Gaussian and nonstationary images.In addition,wavelet transform has been successfully used in the detection of singularity signals recently, including edge detection. The algorithm of edge detection based on wavelet transform is usually more excellent than the traditional ones. But its fault is its neglect of the image directivity. Hereby the algorithm of edge detection based on direction wavelet is studied in this paper. But the width of edge line is too thick. In order to avoid this disadvantage we present a new scheme on the estimate criteria of edge point and on the fusion of the result in several directions. We use the improved algorithm to the detection of small targets. The experiment and theory both indicate that the algorithm of targets detection based on direction wavelet have high precision, strong antinoise and good real-time characteristic.