The fusion analysis of thermal infrared image and visible image is widely used in intelligent security and fault detection. To solve the problem that it is difficult to match two images in the fusion process in the same scene, an image registration method based on featured contour quadrilateral was proposed. Firstly, the thermal infrared image and the visible image were filtered separately by segmentation algorithm. Secondly, the contour points were detected and the contour was redrawn. Then the featured contour was filtered out by the algorithm and the polygon was approximated, and the minimum circumscribed quadrilateral of all contours was generated. The vertexes of this quadrilateral was substituted to calculate the improved homography transformation parameters, and the deviation value of displacement variable was added to neutralize the error of 80%. Finally, the thermal infrared image was registered through the homography transformation model. The experimental results show that this method effectively reduces the matching error, makes the contour positioning more accurate, and can achieve high-precision and rapid image registration.