A New Lateral Cephalogram Image Stitching Technique Using Gaussian Mixture Model and Normalized Cross-Correlation

In dental X-ray lateral cephalogram imaging, linear scanning with a small plate detector is used to obtain sequence images in order to control the radiation dose of the imaging, which requires the image stitching algorithm to stitch it into a complete image. Therefore, this paper proposes a new algorithm based on Gaussian mixture model and normalized cross-correlation to solve the continuous stitching problem of small size X-ray sequence images. The gaussian mixture model is firstly used to segment the high-brightness background area according to the gray information distribution of X-ray image. Then, the target region is used to register the adjacent images, and the genetic algorithm is selected as the optimization algorithm to save the registration time. After the registration parameters between adjacent images are obtained, image fusion is used to make the overlapping area more natural. Finally, after image fusion, the feedback module of gap detection is added to ensure that the registration parameters of each part reach the optimal level and avoid the impact of registration errors. The experimental results show that the algorithm has no stitching errors and gaps, and can effectively solve the stitching problem of small-size X-ray sequence images, it can provide clinical basis for orthodontics treatment.