To improve the fusion effect of infrared and visible light images, a novel fusion method is proposed. The visible light image is segmented into feature subimages with important scene information and grayscale scene subimages using a support vector machine and corrosion expansion algorithm based on the image block. The hot target edge of the infrared image is extracted and enhanced. Target, feature, and gray background subimages are obtained by combining the information of the former feature subimage with the maximum interclass variance method. Two feature subimages(grayscale scene and gray background) are fused by wavelet packet transform. During fusion, different fusion rules are implemented according to the characteristics of the subimages. Additionally, high frequency fusion coefficients are modified to render them more accurate and reliable. The infrared thermal target is injected into the previous fusion result to obtain the final fusion image. Experimental results show that the proposed algorithm is superior to other algorithms both in subjective and objective evaluations.