 This paper proposed a modified histogram equalization with EdgeFusion, MAGF, algorithm to improve the detection rate of maritime infrared target detection in backlight maritime scenes. The algorithm was then extended to include a local contrast saliency models with double scale and modes, LCM-DSM. This algorithm combines two scales and two modes to approach the from bottom to up mechanism of the visual attention model, VM, better and identify the target with a suitable size. Experimental results showed that the algorithm achieved a detection rate of 98.26% under each maritime scene on the average level and could be used in real-time detection with low computational cost. This article was authored by Chang Ding, Zhendong Luo, Yi Fanghou and others.