A comparison of CFAR strategies for blob detection in textured images

Traditional CFAR (constant false alarm rate) approaches applied to the detection of objects in images have proved useful in locating small patches on non-stationary backgrounds. However, the topic of detecting arbitrarily large objects by means of these approaches has received less attention. In this paper we make a comparative analysis of the performance of several CFAR strategies applied to the detection and segmentation of blobs in textured images. The difference in the strategies lies in the way the references for the estimation of the parameters of the detector are considered. By treating four detection schemes through MonteCarlo simulation, we show that directional approaches to the target have better results than non-directional ones. The fourth approach, refered to as ‘gradient-guided’, is the most promising philosophy.