Abstract

In this work, a No-Reference method for assessing quality of ID Card images is developed, using a combination of Face Image Quality Assessment and Text Quality Assessment, motivated by the fact that face and text are two of the main components on an ID Card. For evaluation, a new private dataset was created, consisting of 12,960 Chilean ID Cards, with their corresponding printed facial reference image for face verification. The results obtained with the proposed quality assessment method, show that as more low quality images are discarded, a simultaneous improvement in face and text verification performance metrics is reached.