Optical and SAR image registration has become a research focus in the area of multisensory image processing, because of their information complementarity and feature difference. Based on the structural similarity between images, registration via implicit similarity simplifies the traditional feature matching process as a migration of the feature points and the iterative search of registration parameters only on a single image. This provides a new idea for optical and SAR image registration. Based on the above ideas, this paper uses Canny operator to modify extraction process of feature points; introduces Joint Markov Model (JMM) to improve denoising quality of SAR image; optimizes the search process of registration parameters with modified Quantum Particle Swarm Optimization (QPSO) algorithm; achieves optical and SAR image registration at last. The experiment proves that: improved implicit similarity algorithm on optical and SAR image registration can reach a high accuracy of pixel level or even sub-pixel level.