CM-CAO-C2F Registration Method for Infrared/Visible Light Images Based on Knee Joint Contour

The registration of infrared image with temperature change and visible image has clinical significance. Aiming at the problem that the current diagnosis of knee osteoarthritis is lack of wide applicability and the infrared/visible image with sparse features is difficult to register, a contour based infrared/visible image CM-CAO-C2F registration method of knee osteoarthritis is proposed. After the infrared image and visible image of the knee joint are enhanced, the contour of the legs in the two modal images is extracted; The curvature value of the image contour is calculated and the maximum curvature point is marked. The projection transformation matrix is obtained by coarse to fine registration based on the contour angle direction; The original infrared image is transformed according to the projection transformation matrix to obtain the final registration result. Experimental results show that CM-CAO-C2F achieves better image alignment than CAO-C2F, SIFT, SAR-SIFT, PSO-SIFT; When the image contains less information, the NMRSE values of CM-CAO-C2F are all close to 0, reducing to a certain extent. Compared with SAR-SIFT with smaller NRMSE values, the registration results of $\mathrm{CM}-\mathrm{CAO}-\mathrm{C} 2 \mathrm{~F}$ method are reduced by $35.52 \%$ on average. Compared with the other four methods, the SSIM value of CM-CAO-C2F method is larger. CM-CAO-C2F registration method solves the problem of sparse and difficult registration of infrared/visible image features of the knee joint, and can effectively help doctors to screen and diagnose KOA.