应用自适应模拟退火和多分辨率搜索实现医学超声图像的拼接

为提高图像拼接的成功率,提出了一种基于自适应模拟退火和多分辨率搜索策略的图像自动拼接新方法。该新算法先自适应地选取配准区域,再以互信息为相似度评价标准,结合自适应模拟退火和多分辨率搜索策略的思想分别进行图像平移和旋转参数的全局优化和局部搜索,最后实现图像的拼接。通过对含噪声数字图像和医学超声图像进行的24次模拟拼接实验表明,该新算法较传统的多分辨率直接搜索法有精度高、速度快和抗噪声能力强的优点。由于结合了模拟退火算法的高精度和多分辨率搜索法的高效率,改进后的图像拼接算法将拼接成功率提高了12.5%,并将运算时间控制在可接受的范围内。

Mosaic of medical ultrasound image based on adaptive simulated annealing and multiresolution searching

In order to improve the performance of image mosaic, an automated image mosaic approach was proposed based on adaptive simulated annealing and multiresolution strategy. In the first step of the algorithm, the registration region is adaptively chosen. Then, the parameters of image translation and rotation are optimized based on mutual information with a hybrid method of adaptive simulated annealing for global searching and a multiresolution approach as a local search strategy, and the process of image mosaic is finally achieved. With conducting 24 times of simulated mosaic experiments on both noisy digital images and medical ultrasound images, this algorithm is proved to be a precise, fast and robust mosaic approach compared with the traditional multiresolution direct-search method. Since this improved algorithm inherits the features of high precision from simulated annealing and high efficiency from the multiresolution method, the mosaic performance is improved by 12.5% in terms of success rate while maintaining acceptable speed.