Single Image Deraining Using Bilateral Recurrent Network

Project Code :TCMAPY288

Objective

In this approach we first propose a single recurrent network (SRN) by recursively unfolding a shallow residual network, for extracting rain streaks and also for learning direct mapping for predicting clean background image. Then, we propose bilateral LSTMs, which not only can respectively propagate deep features of rain streak layer and background image layer across stages, but also bring the interplay between these two SRNs, finally forming bilateral recurrent network (BRN).

Abstract

Removing rain streaks plays an important role in many computer vision applications in rainy outdoor scenes, e.g., surveillance, object detection and recognition. Single image deraining is a very challenging. Basically, image deraining can be treated as an image decomposition problem.

To overcome this problem, we propose bilateral LSTMs, which not only can respectively propagate deep features of rain streak layer and background image layer across stages, but also bring the interplay between these two SRNs, finally forming bilateral recurrent network (BRN).

The proposed methods also perform more favourably in terms of generalization performance on real-world rainy dataset. We also proposed a simple yet effective single recurrent network (i.e., SRN) for image deraining.

Keywords: Image De-raining, Convolutional Neural Network, Recurrent Network, LSTM.

NOTE: Without the concern of our team, please don't submit to the college. This Abstract varies based on student requirements.