基于方向小波字典的相关跟踪图像特征识别和定位方法研究

Image stabilization control is one of the key technologies for the large-aperture high-resolution space solar telescopes and space cameras. Add image stabilization control system in satellite, can achieve high quality imaging with high resolution. This project researches the image feature recognition and location in correlation trackers with image spares representation. Using a Gabor wavelet frame, a steerable wavelet frame and a curvelet frame constitute a directional wavelet dictionary. Solar images are decomposed on the directional wavelet dictionary in order to obtain the sparse representation of the typical structure，such as granulations，sunspots, pores and faculae, in solar images. Based on this sparse representation ，we expect to find feature quantities describing these structures so that the tracked target image can be selected from the large field of view images by correlation trackers. In order to find effective to describe the symmetry feature, we study the relationship between the symmetry of the horizontal and vertical image offset calculation accuracy and the symmetry of the image structure using directional wavelets. The fast tracked target image selection algorithm and its hardware and software implementation are researched. The validity of this fast algorithm is verified using the correlation tracker test system. The research results can be used to improve the performance of correlation trackers and acquire the high-resolution solar images.

图像稳定控制技术是大口径高分辨率空间太阳望远镜和空间相机中不可缺少的关键技术之一，在卫星中加入图像稳定控制系统，可以实现高分辨率和高质量成像。本项目采用图像稀疏表示方法研究相关跟踪器中的图像特征识别和定位。用Gabor wavelet框架、Steerable wavelet框架和Curvelet框架构成方向小波字典，用此字典分解太阳图像，以期望获得对太阳图像中的典型结构，如米粒、黑子、气孔和亮斑等的稀疏表示，在此基础上找到描述这些典型结构的特征量，以使相关跟踪器能够从大视场图像中选出适合相关跟踪的目标图像。本项目还将利用方向小波研究图像结构的对称性与水平和垂直方向图像偏移量计算精度对称性间的关系，以寻找有效表示对称性的特征量。研究跟踪目标图像选择的快速算法及其软硬件实现，并在相关跟踪器实验系统上对算法的有效性进行验证。研究成果对提高相关跟踪器的性能和获取高分辩率太阳图像有重要意义。