Computer Science > Computer Vision and Pattern Recognition
[Submitted on 7 May 2021]
Title:Toward Interactive Modulation for Photo-Realistic Image Restoration
View PDFAbstract:Modulating image restoration level aims to generate a restored image by altering a factor that represents the restoration strength. Previous works mainly focused on optimizing the mean squared reconstruction error, which brings high reconstruction accuracy but lacks finer texture details. This paper presents a Controllable Unet Generative Adversarial Network (CUGAN) to generate high-frequency textures in the modulation tasks. CUGAN consists of two modules -- base networks and condition networks. The base networks comprise a generator and a discriminator. In the generator, we realize the interactive control of restoration levels by tuning the weights of different features from different scales in the Unet architecture. Moreover, we adaptively modulate the intermediate features in the discriminator according to the severity of degradations. The condition networks accept the condition vector (encoded degradation information) as input, then generate modulation parameters for both the generator and the discriminator. During testing, users can control the output effects by tweaking the condition vector. We also provide a smooth transition between GAN and MSE effects by a simple transition method. Extensive experiments demonstrate that the proposed CUGAN achieves excellent performance on image restoration modulation tasks.
Current browse context:
cs.CV
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.