본문 바로가기
HOME> 논문 > 논문 검색상세

학위논문 상세정보

Adaptive single image dehazing 원문보기

  • 저자

    Bui Minh Trung

  • 학위수여기관

    경희대학교

  • 학위구분

    국내석사

  • 학과

    전자·전파공학과

  • 지도교수

    Wonha Kim

  • 발행년도

    2014

  • 총페이지

    vi, 31 p.

  • 키워드

  • 언어

    eng

  • 원문 URL

    http://www.riss.kr/link?id=T13536590&outLink=K  

  • 초록

    Goal of this dissertation is to enhance visual quality of hazy image and video for satisfying human perception while sustaining low computational complexity. Based on observation that thickness of haze is proportional to local brightness and color variance, and it also smoothly varies within objects, but suddenly changes at edges, we proposed a simple but effective prior - edge-preserving smooth dark channel prior (ESDP) for approximating haze thickness from a single image. We then use this prior in directly estimating airlight and transmission without adopting any post-refinements as conventional methods, and consequently these estimated values are used to recover high-quality haze-free image. We also extend our dehazing methods in video dehazing. By observation that abrupt color transition between consecutively dehazed frames causes flickers in video dehazing, we proposed a framework of temporal interpolation, which results the smooth color transition while preventing potential blur by motion difference. The experiment verifies that the proposed method significantly enhance visual quality of hazy image and video while requiring low computational complexity.


 활용도 분석

  • 상세보기

    amChart 영역
  • 원문보기

    amChart 영역