PROJECT TITLE :

Efficient Evaluation of Image Quality via Deep-Learning Approximation of Perceptual Metrics

ABSTRACT:

An important role in the evaluation of complicated Image Processing algorithms is played by image metrics based on human visual system (HVS). As a result of this, the use of the HVS model is restricted to a small number of applications and a small amount of input data. For all of these reasons, real-world settings do not favour such measurements. Deep Image Quality Metric (DIQM), a deep-learning approach to learn the global image quality feature, is proposed to address these challenges (mean-opinion-score). When compared to prior solutions, DIQM is able to efficiently imitate existing visual metrics while reducing computing costs by an order of magnitude.