CFAR Method Based on Half Side Gaussian Model for Small Target Detection

This work deals with the problem of small target detection in infrared image. A new CFAR method based on half side Gaussian distribution is presented for the segmentation of morphological filtered infrared image. First, the distribution of the residual image after preprocessing based on morphological filters is exploited. Although widely used and useful, the traditional Gaussian distribution does not appear to be the best choice for modeling the residual images. Half side Gaussian distribution (HG) is adopted to model the residuals since it fits the data better. Then, based on the HG model a new CFAR (constant false alarm rate) threshold method called HG-CFAR is proposed. At last, comparisons are made between HG-CFAR and traditional CFAR based on Gaussian distribution. The results show that new method is more efficient.