ASKI: full-sky lensing map-making algorithms
ASKI: full-sky lensing map-making algorithms
复制标题
DOI：
10.1111/j.1365-2966.2009.15609.x
复制
发表时间：
2009-01
影响因子：
4.8
通讯作者：
C. Pichon;É. Thiébaut;S. Prunet;K. Benabed;S. Colombi;T. Sousbie;R. Teyssier
中科院分区：
文献类型：
--
作者：
C. Pichon;É. Thiébaut;S. Prunet;K. Benabed;S. Colombi;T. Sousbie;R. Teyssier
在即将到来的全天空透镜巡天的背景下，提出了边缘保持非线性算法ASKI（全天空κ反演）。使用最大后验概率反演的框架，它的目的是恢复最佳的全天空收敛图从噪声调查与面具。ASKI提供了两个步骤：（一）CCD图像的可能拥挤的银河领域去模糊使用自动边缘保持反卷积;（二）一旦减少剪切估计使用标准技术，部分掩盖的收敛图也通过边缘保持方法反转。图像去模糊的效率通过减小的剪切的质量因子中的相对增益来量化，如SEXTRACTOR所估计的。交叉验证作为删除的恒星数量的函数，产生一个自动估计的最佳水平的正规化的星系的反卷积。据发现，当所观察到的领域是拥挤的，这种增益可以是相当显着的现实陆基8米级调查。当在迭代反卷积期间施加正性和边缘保持l1 - l2惩罚时，发生最显著的改进。在HORIZON-4π N-体模拟的信噪比（S/N）在lcut = 500-2500范围内的噪声图上研究了收敛反演的质量，并使用单点统计量（S3和S4）、功率谱、簇计数、峰值补丁和骨架进行了量化。结果表明：（i）重建算法能够在银河系切割/非均匀噪声中进行插值和外推;（ii）其保留锐度的惩罚避免了地图簇附近的强偏置;（iii）它很好地重建了由偏度和峰度跟踪的PDF形状;（iv）重建地图的几何和拓扑结构与初始地图接近，如由峰值斑块分布和骨架的差分长度所追踪的;（v）恢复的地图的两点统计与初始地图的对应平滑版本一致;（vi）点源的分布也与对应平滑一致，当应用L1 - L2先验时具有显著改进。污染的B模式时，现实的银河切割存在也进行了调查。泄漏主要发生在大范围内。在该模型中实现的非线性在该领域的峰值附近的小尺度上是显着的。
Within the context of upcoming full-sky lensing surveys, the edge-preserving non-linear algorithm ASKI (All-Sky κ Inversion) is presented. Using the framework of Maximum A Posteriori inversion, it aims at recovering the optimal full-sky convergence map from noisy surveys with masks. ASKI contributes two steps: (i) CCD images of possibly crowded galactic fields are deblurred using automated edge-preserving deconvolution; (ii) once the reduced shear is estimated using standard techniques, the partially masked convergence map is also inverted via an edge-preserving method. The efficiency of the deblurring of the image is quantified by the relative gain in the quality factor of the reduced shear, as estimated by SEXTRACTOR. Cross-validation as a function of the number of stars removed yields an automatic estimate of the optimal level of regularization for the deconvolution of the galaxies. It is found that when the observed field is crowded, this gain can be quite significant for realistic ground-based 8-m class surveys. The most significant improvement occurs when both positivity and edge-preserving l1 - l2 penalties are imposed during the iterative deconvolution. The quality of the convergence inversion is investigated on noisy maps derived from the HORIZON-4π N-body simulation with a signal-to-noise ratio (S/N) within the range lcut = 500-2500, with and without Galactic cuts, and quantified using one-point statistics (S3 and S4), power spectra, cluster counts, peak patches and the skeleton. It is found that (i) the reconstruction is able to interpolate and extrapolate within the Galactic cuts/non-uniform noise; (ii) its sharpness-preserving penalization avoids strong biasing near the clusters of the map; (iii) it reconstructs well the shape of the PDF as traced by its skewness and kurtosis; (iv) the geometry and topology of the reconstructed map are close to the initial map as traced by the peak patch distribution and the skeleton's differential length; (v) the two-point statistics of the recovered map are consistent with the corresponding smoothed version of the initial map; (vi) the distribution of point sources is also consistent with the corresponding smoothing, with a significant improvement when l1 - l2 prior is applied. The contamination of B modes when realistic Galactic cuts are present is also investigated. Leakage mainly occurs on large scales. The non-linearities implemented in the model are significant on small scales near the peaks in the field.