采用目标区域互信息的星空图像配准

Stellar image registration based on mutual information in object area

提出了采用目标区域互信息的测度方法对星图进行精确配准以解决星图中存在噪声、伪星点、星点稀疏以及星图间的旋转等问题。首先对星图进行图像分割,检测出星点目标并对星点进行二值化处理;然后基于互信息配准模型,在含星点的目标区域上,利用Powell算法将最大互信息作为目标函数来指导图像间最优变换参数的搜索。分析了适宜于互信息测度配准的星点分割算法,并论证了采用目标区域互信息的星图配准的可行性。对提出的算法与标准的互信息配准算法进行了对比。结果表明:提出算法的时间消耗与图像中星的数量有关,在图像大小为1 000×1 000时,提出算法的加速比为标准算法的3.4倍。该算法在星图中存在噪声、伪星点、星点稀疏和旋转的情况下仍能进行准确配准,50组实拍星图配准误差平均值为0.138 2pixel,满足了星空图像对精确配准的要求。

A measuring algorithm based on mutual information in an object area was proposed to register stellar images with noises, pseudo stars, sparse stars and the rotation between the images. Firstly, a stellar image was segmented to extract star points in the stellar image and the star points were processed with binarization. Then, on the basis of the mutual information model, the Powell algorithm was used to guide the search of the best transformation parameters in the object area by taking the maximal mutual information as the aim function. Furthermore, the star segmentation method suitable for the proposed stellar registration algorithm was analyzed and the feasibility of the proposed algorithm was verified. The proposed algorithm was compared with normal mutual information registration algorithm , and the results show that the time consuming is relative to the number of stars in image, and the speed up ratio is up to 3.4 times that of the normal mutual information registration algorithm when the image size is 1 000 × 1 000. Experimental results demonstrate that the proposed algorithm achieves a high precision stellar registration with noises, pseudo stars, sparse stars and rotation. The average error of 50 groups of real stellar images is 0. 1382 pixel. It concludes that the algorithm meets the requirements of space target detection for image registration.