<title>CFAR-enhanced image codec for SONAR ATR</title>

Abstract

The wavelet transform gives a compact multiscale representation of a digital image and provides a hierarchical structure which is well suited for post- processing. With its good localization property in both the spatial domain and the frequency domain, wavelet-based image compression has gained a huge success in the past several years. In this paper we present a wavelet-based image codec specially designed for an automatic target recognition (ATR) system. Due to the large amount of data size, a `good' image compression algorithm is both necessary and important as a pre-processing stage for an ATR system, especially in a `real-time' processing situation. We incorporate a constant false alarm rate (CFAR) detector into an embedded image compression algorithm to efficiently code target pixels exclusively in the bit stream. The new image codec, which is enhanced by the CFAR feature, clearly exhibits (potential) targets in the decompressed image. Another new feature of this codec is a wavelet-interpretation of the CFAR detector, a multiscale representation of the CFAR values in the wavelet domain.