A Composite Kernel Regression Method Integrating Spatial and Gray Information for Infrared Small Target Detection

Small target detection in infrared imagery with complex background is always an important task in infrared target tracking system. Complex clutter background usually results in serious false alarm because of low contrast of infrared imagery. In this paper, a composite kernel regression method is proposed for infrared small target detection. In the proposed method, a nonlinear regression model is firstly built based on a multiplicatively-composite kernel which integrates both spatial and gray information surrounding interesting pixels. Then the composite kernel regression is utilized to estimate the clutter background of image. At last, two-parameter CFAR detection is performed on background-removed infrared image to extract the target. Experimental results prove that the proposed algorithm is effective and adaptable to small target detection with complex background.