Regularization via deep generative models: an analysis point of view
This addresses regularization for imaging problems, offering a more versatile and robust method, though it is incremental as it builds on existing generative model frameworks.
The paper tackles the problem of regularizing inverse imaging tasks like inpainting, deblurring, and super-resolution by using a deep generative model, proposing an analysis formulation that directly optimizes the image and penalizes the latent vector, which leads to improved performance and robustness compared to previous synthesis approaches.
This paper proposes a new way of regularizing an inverse problem in imaging (e.g., deblurring or inpainting) by means of a deep generative neural network. Compared to end-to-end models, such approaches seem particularly interesting since the same network can be used for many different problems and experimental conditions, as soon as the generative model is suited to the data. Previous works proposed to use a synthesis framework, where the estimation is performed on the latent vector, the solution being obtained afterwards via the decoder. Instead, we propose an analysis formulation where we directly optimize the image itself and penalize the latent vector. We illustrate the interest of such a formulation by running experiments of inpainting, deblurring and super-resolution. In many cases our technique achieves a clear improvement of the performance and seems to be more robust, in particular with respect to initialization.