Abstract: Aiming at the single character of target image acquisition in the traditional bionic vision system, multispectral visual image processing method based on adaptive regulation of humanoid eye was proposed. Firstly, the improved automatic focusing algorithm was used to collect the high resolution image of visible light and the low resolution image of near-infrared light. In the multispectral imaging system, there were different problems of visible and near-infrared image resolution under fixed focal length due to different refractive index of spectral prism. The improved two-generation wavelet transform was adopted to enhance the image contrast and improve the visual effect. Finally, the performance of automatic focus algorithm and image enhancement algorithm was verified by using a multi-spectral experimental device based on liquid zoom lens. The experimental results indicate that the average time of effective auto focusing system is 756 ms, and the gray variance function value increases by 79.4% after near infrared image enhancement, which solves the problem of low contrast and fuzzy details and realizes adaptive regulation.