最大熵与MP-CLEAN方法对扩展源图像重建的比较  

A Comparison of the MEM and the MP-CLEAN Methods for the Image Reconstruction of Extended Sources with Simulated Visibility Data

作　　者：张茂林[1,2] 周建锋[1,2] 

机构地区：[1]清华大学工程物理系天体物理中心,北京100084 [2]粒子技术与辐射成像教育部重点实验室(清华大学),北京100084

出　　处：《天文研究与技术》2016年第1期100-110,共11页Astronomical Research & Technology

基　　金：国家自然科学基金(11173038;11373025);清华大学自主科研计划(20111081102)资助

摘　　要：在射电综合成像领域,通常需要用退卷积的方法补全频域稀疏的采样。由于扩展源的频域信息更为丰富,要补全这些信息相对于点源来说更难,因此扩展源的图像重建是射电综合成像领域的一大难点。为了探索射电干涉扩展源图像重建方法的特点,将最大熵与一种加速的CLEAN方法(文中称之为Multi-Point CLEAN,MP-CLEAN)对扩展源的干涉阵模拟数据的图像重建进行了比较。通过比较,发现对于同样的观测数据,两种方法都能较好地重建图像,但MP-CLEAN方法的旁瓣祛除效果与重建效果优于最大熵方法,而且在模拟数据重建中MP-CLEAN方法的总体速度比最大熵快3倍以上。最后,在讨论部分通过研究两种方法中参数的选择对重建结果的影响,发现最大熵方法比MP-CLEAN方法对参数选择的依赖性弱,这表明最大熵方法有更好的鲁棒性。In the radio synthesis imaging area, the deconvolution is used to fill up the un-sampled points at the Fourier domain. The algorithms like CLEAN and MEM often play an important role in the deconvolution. As extended sources have more complex visibility, the reconstruction for extended sourcesis difficult. In this paper we make a comparison of the MEM （ Maximum Entropy Method） with the speed up CLEAN method （named Multi-Point CLEAN in this paper, or MP-CLEAN for short） for extended source reconstruction based on the simulated visibility data. The simulated data are produced by sampling the Flourier Frequency domain for a true image with the sample function generated from a simulated interferometer array. By analyzing and comparing, it is easy to find that both of these two methods work well on the image reconstruction while the MP-CLEAN is better than the MEM for getting rid of the side lobes. The MP-CLEAN is at least 3 times faster than the MEM in our simulations. A deeper analyzing of the selection of parameters shows that the MEM is more independent on the parameters than the MP-CLEAN, which indicates that the MEM is more robust.

关 键 词：图像重建 射电干涉阵 Multi-PointCLEAN 最大熵方法 扩展源 

分 类 号：P164[天文地球—天文学]