Learning Dual Convolutional Neural Networks for Low-Level Vision

TLDR

Low‑level vision tasks such as super‑resolution, edge‑preserving filtering, deraining, and dehazing typically require estimating both structures and details of the target signals. The authors propose a DualCNN that uses two parallel branches to recover structures and details for low‑level vision tasks. DualCNN reconstructs target signals by combining recovered structures and details according to each task’s formation model, and its flexible design allows easy integration into existing CNNs. Experiments demonstrate that DualCNN achieves state‑of‑the‑art performance on multiple low‑level vision tasks.

Abstract

In this paper, we propose a general dual convolutional neural network (DualCNN) for low-level vision problems, e.g., super-resolution, edge-preserving filtering, deraining and dehazing. These problems usually involve the estimation of two components of the target signals: structures and details. Motivated by this, our proposed DualCNN consists of two parallel branches, which respectively recovers the structures and details in an end-to-end manner. The recovered structures and details can generate the target signals according to the formation model for each particular application. The DualCNN is a flexible framework for low-level vision tasks and can be easily incorporated into existing CNNs. Experimental results show that the DualCNN can be effectively applied to numerous low-level vision tasks with favorable performance against the state-of-the-art methods.