Date&Time : 

2020 / 12 / 18 (Fri) 14:20 ~ 16:20

 

Location : 

Delta Building R216

 

Speaker : 

康立威 教授 ( 國立臺灣師範大學電機工程學系)

 

Topic : 

Image Restoration: From Sparse Coding to Deep Learning (影像品質回復技術: 從稀疏編碼到深度學習)

 

 

Abstract : 

Image restoration is a classical problem in image processing and computer vision community. The problem is challenge and ill-posed. Based on the fact that images in the real world may usually suffer from possible noises of different types, and the effectiveness of the related vision-based applications may be degraded accordingly. Therefore, it is important to restore the degraded images to the corresponding acceptable visual qualities. In this talk, I will focus on the restoration of bad weather images, such as rainy or hazy images. I will first introduce our contributions to bad weather image restoration via sparse coding relying on the intrinsic sparse property of images, including the first single image-based rain streaks removal framework. The sparse coding framework has been also shown to be applicable to Gaussian noise removal and single image super-solution. Recently, relying on the rapid development of deep learning techniques with great success in numerous perceptual tasks, several deep learning-based bad weather image restoration methods have been also presented. In addition, sparse coding and deep learning have also been shown to be with certain relation. I will also present our recent contributions to single image haze removal and rain removal via deep learning, which have been shown to achieve state-of-the-art performances. I will also briefly introduce some our on-going works in other types of single image denoising and discuss possibly future research directions.

 

Short Bio :

現職: 國立臺灣師範大學電機工程系副教授

經歷: 國立雲林科技大學資訊工程系副教授、助理教授、中央研究院資訊科學研究所助理研究學者、博士後研究員

學歷: 國立中正大學資訊工程系博士、碩士、學士

研究領域: 電腦視覺、影像處理、多媒體訊號處理、深度學習