Summary

Smart Info-Media Systems in Asia

2019

Session Number:RS3

Session:

Number:RS3-1

Fast Haze Removal of Fixed-point Camera Images Using Temporal Update of Transmission Map

Tatsuya Hamamura,  Tomoaki Kimura,  Hiroyuki Tsuji,  

pp.131-135

Publication Date:2019/9/4

Online ISSN:2188-5079

DOI:10.34385/proc.57.RS3-1

PDF download (1.6MB)

Summary:
Outdoor images are often degraded by atmosphericphenomena such as haze, fog, smoke, etc. The algorithm of recovering contrast and color fidelity of such images is called haze removal. The purpose of this paper is to propose fast haze removal algorithm designed for the time-series images captured by a fixed point camera such as surveillance. In the proposed method, the difference information of successive two images is used for updating the transmission map, although only the first image should be de-hazed by using the conventional single-image haze removal algorithm proposed by Kaiming He, et al., or dark channel prior. The advantage of our method is to reduce the processing time of transmission map generation significantly, preventing deterioration of the resulted images compared with the iterative application of He’s conventional method to each video frame. It is verified through several experiments that the proposed temporal update is approximately 10 times faster than the conventional transmission map generation using guided image filters.