A robust adaptive multi-spectral object detection by using wavelet transform

Abstract

A robust multi-spectral adaptive object detection algorithm is derived for a multiple spatial resolution and orientation decomposition image mode by using a wavelet transform. The second-order statistical moments of the wavelet transform are computed for a random image field and then used to develop a generalized maximum likelihood ratio test and to analyze detection performance. The computational cost of the new detector can be reduced substantially when compared to conventional spatial size and orientation matched filter-bank approach by using a coarse-to-fine spatial matching strategy. >