Report on the performance of image differencing from the perspective of the learning-based classifier task

Abstract

The output of the image differencing task (subtractImages combined with detectAndMeasure) is the input to the learning based classification module (rbClassifier) in AP pipelines. Therefore, the quality of the generated difference images and the detected sources, is of a high importance from the classifier’s point of view. This tech-note tries to provide a basic statistical analysis of this data product as generated with the current version of the AP pipeline. Obviously, the classifier itself is not involved neither in the pipelines nor the analysis provided in this report.