Infrared Dim and Small Target Detection Based on the Human Visual Attention Mechanism

This article proposes an algorithm based on the human visual attention mechanism to solve the infrared target detection problem provided that the targets are submerged in the background. Firstly, the regions of interest (ROIs) of the image are obtained by the top-hat transform of mathematical morphology and the method of adaptive thresholding segmentation. Secondly, the image’s signal-to-noise ratio (SNR) is enhanced by processing the ROIs using the difference of Gaussians (Dog) filter which has the characteristic of human vision in the scale space. Then, the points which have the local maximum SNR in the detected image can be regarded as the candidate targets. At last, considering the targets are easily submerged in the background and to prevent the targets not being detected, the algorithm proposes searching the missing targets again using the Retinex theory. Experimental results with real forward-looking infrared (FLIR) images show higher detection rate and lower false alarm rate than other methods, especially for the targets submerged in the background.