ExCNet: A Framework for Zero-Shot Restoration of Back-lit Images Using Deep Internal Learning

School of Software Engineering, Tongji University, Shanghai, China


Introduction

This is the website for our work "Zero-Shot Restoration of Back-lit Images Using Deep Internal Learning".

How to restore back-lit images still remains a challenging task. State-of-the-art methods in this field are based on supervised learning and thus they are usually restricted to specific training data. In this paper, we propose a "zero-shot" scheme for back-lit image restoration, which exploits the power of deep learning, but does not rely on any prior image examples or prior training. Specifically, we train a small image-specific CNN, namely ExCNet (short for Exposure Correction Network) at test time, to estimate the ``S-curve'' that best fits the test back-lit image. Once the S-curve is estimated, the test image can be then restored straightforwardly. ExCNet can adapt itself to different settings per image. This makes our approach widely applicable to different shooting scenes and kinds of back-lighting conditions. Statistical studies performed on 1512 real back-lit images demonstrate that our approach can outperform the competitors by a large margin. To the best of our knowledge, our scheme is the first unsupervised CNN-based back-lit image restoration method. To make the results reproducible, the source code is available on this website.


Key Ideas of ExCNet

(1) The core of our approach is a specially designed CNN, namely ExCNet (Exposure Correction Network). Given a test back-lit image I, ExCNet can be trained in an image specific way to estimate the parametric "S-curve" that best fits I. S-curve is widely adopted by photo editing softwares as an interactive tool for manually correcting ill-exposed images. When S-curve is ready, I can be restored straightforwardly. To our knowledge, our work is the first unsupervised learning framework to correct image exposure automatically. Our image-specific method can be widely applicable to different shooting scenes and kinds of lighting conditions.

(2) When designing ExCNet, a key issue is how to devise a loss function that can evaluate an image's degree of being ill-exposed. To this end, motivated by the formulation of MRF (Markov Random Field), we design a block-based loss function, which tends to maximize the visibility of all blocks while keeping the relative difference between neighboring blocks.

(3) Due to the CNN structure of ExCNet, our method could learn the mapping relationship between images and their best "S-curve" parameters. Thus along with the increasing of processed images, ExCNet takes less iterations to converge to the optimized curve when facing an unseen image. Even though it is trained at test time, it runs very fast since the network is small and the training usually converges after several iterations. Besides, when handling a video stream, the correction of subsequent frames could be guided by the parameters of the previous frames, which would not lead to significant flickering artifacts and have a relatively low computational cost. Compared with typical supervised learning based schemes, ExCNet is more suitable to be implemented in a mobile or on-chip system.

The structure of ExCNet. Each training iteration of ExCNet can be conceptually considered as having two stages, adjusting Il using the intermediate S-curve and deriving the loss.

The overall pipeline of back-lit image restoration using ExCNet. The restoration pipeline comprises three major steps, S-curve estimation, luminance channel restoration, and chrominance adjustment.


Test examples
In order to facilitate readers to visually compare the results of different back-lit image restoration approaches, we show some examples.


Source Codes

1. ExCNet.ipynb

This is the code of ExCNet, including the network model and the back-lit images restoration procedure. The prerequisites for running the .ipynb file is the Tensorflow environment and Jupyter Notebook.

2. BacklitImages.zip, BacklitVideos.zip

These are some back-lit images and videos for testing. The extract codes are 'v752' and 'yerr' respectively.


Last update: Sep. 11, 2019