High Impact Factor : 4.396 icon | Submit Manuscript Online icon |

Learning to See in the Dark

Author(s):

Ajay Kumar Maurya , Dr. Ambedkar Institute of Technology for Handicapped kanpur; Abhinav Mishra, Dr. Ambedkar Institute of Technology for Handicapped kanpur; Aniket Gupta, Dr. Ambedkar Institute of Technology for Handicapped kanpur

Keywords:

SNR, CNN, CAN, Exposure Time, U-Net, Deep Neural Network

Abstract

Low light imaging is a challenge of cameras because of low photon counts and low signal to noise ratio. In case of low lighting taking an image becomes more challenging and complex. Generally all camera supports low light imaging but not all images result are good because they mostly uses traditional pipeline. Traditional Pipelining performs various processing such as Histogram Processing or scaling but it does not resolve the low SNR because of low photon counts. And generally short exposure images suffers from noise, while long exposure images may induce blur due to camera shake or Object motion and is often impractical. There are many different types of techniques like denoising, deblurring and image enhancement which are already proposed but all these techniques have a certain limits that the images are take in dim environment but they cannot give better result in extremely low light. In this paper we have proposed a learning based pipeline which uses Artificial intelligence to train his neurons to improve the learning, which can see in extremely low lighting. In this paper we are introducing a dataset of raw short-exposure low light images to correspond with images of long exposure time.

Other Details

Paper ID: IJSRDV7I40323
Published in: Volume : 7, Issue : 4
Publication Date: 01/07/2019
Page(s): 630-633

Article Preview

Download Article