DNF – Galaxy photometric redshift by Directional Neighbourhood Fitting
Abstract
Wide field images taken in several photometric bands allow simultaneous measurement of redshifts for thousands of galaxies. A variety of algorithms to make this measurement have appeared in the last few years, the majority of which can be classified as either template- or training-based methods. Among the latter, nearest neighbour estimators stand out as one of the most successful, in terms of both precision and the quality of error estimation. In this paper we describe the Directional Neighbourhood Fitting (DNF) algorithm based on the following: a new neighbourhood metric (Directional Neighbourhood), a photo-z estimation strategy (Neighbourhood Fitting) and a method for generating the photo-z probability distribution function. We compare DNF with other well-known empirical photometric redshift tools using different public data sets (Sloan Digital Sky Survey, VIMOS VLT Deep Survey and Photo-z Accuracy Testing). DNF achieves high-quality results with reliable error.