Canny Algorithm: A New Estimator for Primordial Non-Gaussianities

Abstract

We utilize the Canny edge detection algorithm as an estimator for primordial non-Gaussianities. In preliminary tests on simulated sky patches with a window size of 57 degrees and multipole moments $l$ up to 1024, we find a $3\sigma$ distinction between maps with local non-Gaussianity $f_{NL}=350$ (or $f_{NL}=-700$) and Gaussian maps. We present evidence that high resolution CMB studies will strongly enhance the sensitivity of the Canny algorithm to non-Gaussianity, making it a promising technique to estimate primordial non-Gaussianity.

Comments: 4 pages, 4 figures; v2. 5pp, as submitted to PRD; v3. 5pp, minor clarifications and added discussion of negative fNL values