Jaakko T. Astola, Karen O. Egiazarian, Igor V. Kurbatov, Alexander V. Totsky, and Alexander A. Zelensky

In this paper, we propose bispectrum based technique for unknown object reconstruction and recognition from observed jittery and noisy 2-D image realizations (frames), when only a small number of frames are disposal. The technique is based on two following ideas. The first one is the jitter removal by derivation the differences between coordinates of maximums of jittery and noisy neighbour 1-D line image cross-correlations and centre of gravity (CG) values computed by bispectrum recovery of neighbour 1-D line image cross-correlations. The second one is the introduction the pre-distortions to every line image to avoid on one hand phase unwrapping and phase ambiguity and on the other hand to stabilize CG values of the pre-distorted 1-D line images to obtain reliable alignment in 2-D bispectrum reconstructed image. Computer simulations have been provided to illustrate the efficiency of the proposed approach. The performance of the proposed technique is studied by visual examination of the reconstructed test images. It is shown that the proposed technique can effectively remove heavy jitter distortions in the background of quite intensive additive Gaussian noise to recognize unknown object reliably.