Many image restoration problems are ill-posed in nature, hence, beyond the input image,
most existing methods rely on a carefully engineered image prior, which enforces some
local image consistency in the recovered image. How tightly the prior assumptions are
fulfilled has a big impact on the resulting task performance. To obtain more flexibility, in this
work, we proposed to design the image prior in a data-driven manner. Instead of explicitly
defining the prior, we learn it using deep generative models. We demonstrate that this …

Image restoration tasks, such as deblurring, or super-resolution, can be cast as inverse
problems, where we seek to retrieve a clean image from a degraded observation. In order to
determine how to recreate the missing information in the degraded observation, it is
necessary to define a model of the properties of the expected solution. From a Bayesian
perspective, this model of the solution is defined as the prior, and solving the inverse
problem amounts to finding an image that provides a good compromise between the prior …