How to Optimize the Utilization of Image Quality Metrics in Computer Vision?
Aladine Chetouani, Mohammed El Hassouni · 2018
In this paper, we propose to show the importance to consider the image quality in Computer Vision (CV) applications. We also describe a proposed framework that not only take into account the quality but rather permits to select the more adapted measure for a given CV application. Here, the selection of the image quality metric is based on a degradation identification step using a Linear Discriminant Analysis (LDA) method. The proposed framework has been applied to a Full-Reference approach where the reference image is supposed to be available and for No-Reference approach where only the captured image is accessible. The method has been tested using the TID 2008 database, which is composed of 17 degradation types.