A new automatic target recognition algorithm based on infrared multispectral imagery correlation (IMIC) is presented. According to infrared multispectral characteristic information (such as radiation intensity and spectral distribution) of targets, background and decoys, an infrared multispectral eigenmatrix of scenario is constructed. An infrared multispectral eigenmatrix of interesting targets is reconstructed after segmenting image by maximum-error-method and fusing target information of space and spectrum. An infrared multispectral imagery correlation recognition method is formulated in terms of spectral radiation difference of interesting targets. Digital simulation experiment result verifies that the algorithm is proper and feasible.