Infrared small dim target detection based on local contrast combined with region saliency

A part local contrast measure algorithm is proposed to solve the problem of low efficiency of original local contrast measure, which combines region saliency with original local contrast measure. Instead of finding target in the whole image, the local contrast measure is constrained in saliency region of current image frame in the proposed method. At the first stage, the image entropy and local similarity feature are used to evaluate the saliency of infrared images, which measures the saliency region of a single frame. At the second stage, the local contrast measure is presented in saliency region, which forms the part local contrast map. An adaptive threshold is adopted to segment the target from part local contrast map. Experiments on several real infrared image sequences have validated the ability of the proposed method in improving the signal-to-noise ratio and the detection capability. In particular, the proposed method can reduce false alarm rate, which is an inherent defect of original local contrast method. The high efficiency is also an important strength of the proposed method, which leads to wide application prospect.