Learning to Predict Image‐based Rendering Artifacts with Respect to a Hidden Reference Image
Learning to Predict Image‐based Rendering Artifacts with Respect to a Hidden Reference Image
Abstract Image metrics predict the perceived per‐pixel difference between a reference image and its degraded (e. g., re‐rendered) version. In several important applications, the reference image is not available and image metrics cannot be applied. We devise a neural network architecture and training procedure that allows predicting the MSE, SSIM …