The existed infrared intensity and polarization image fusion algorithm could not dynamically adjust fusion algorithm according to the change of image difference characteristics,which results in that the partial differences feature fusion effect was not ideal or even failure.According to thought of mimicry bionics and learning from Mimicry Octopus′s multi-mimicry process, a fusion method of combination fusion of multi-types mimic variable of infrared intensity and polarization images was proposed. Firstly, Mimicry Ooctopus′s multi-mimicry process,the reason of multi-mimicry were analyzed. Secondly, the correspondence between multi-mimicry process and image fusion process was found, and multiple types variables of the image fusion process were determined.Finally, a reversible variable composition relation for image fusion was established and image fusion was carried out that used the combined relation. The experimental results showed that the information entropy, standard deviation, edge intensity, average gradient and sharpness of the obtained fusion image were obviously improved 1.16%, 7.25%, 3.00%, 0.31%, 10.18% in average. The establishment of this method could make the fusion algorithm within the variable selection and combination according to the original image difference characteristics of the dynamic adjustment, so as to obtain a targeted fusion algorithm.