BP-DIP: A Backprojection based Deep Image Prior
This is an incremental improvement for image restoration tasks, addressing data scarcity and model mismatch issues in computer vision.
The paper tackled the problem of deep learning methods for image restoration requiring large training datasets and suffering from performance drops when test conditions mismatch training, by combining Deep Image Prior with a backprojection fidelity term, resulting in higher PSNR values and better inference run-time for deblurring compared to plain DIP.
Deep neural networks are a very powerful tool for many computer vision tasks, including image restoration, exhibiting state-of-the-art results. However, the performance of deep learning methods tends to drop once the observation model used in training mismatches the one in test time. In addition, most deep learning methods require vast amounts of training data, which are not accessible in many applications. To mitigate these disadvantages, we propose to combine two image restoration approaches: (i) Deep Image Prior (DIP), which trains a convolutional neural network (CNN) from scratch in test time using the given degraded image. It does not require any training data and builds on the implicit prior imposed by the CNN architecture; and (ii) a backprojection (BP) fidelity term, which is an alternative to the standard least squares loss that is usually used in previous DIP works. We demonstrate the performance of the proposed method, termed BP-DIP, on the deblurring task and show its advantages over the plain DIP, with both higher PSNR values and better inference run-time.