single image haze removal using dark channel prior

36
Single Image Haze Removal Using Dark Channel Prior Professor : 王王王 王王 Student : 王王王 CVPR 2009 . Best Paper Award Kaiming He , Dept. of Inf. Eng., Chinese Un iv. of Hong Kong, Hong Kong, China Jian Sun Xiaoou Tang , Dept. of Inf. Eng., Chinese U niv. of Hong Kong, Hong Kong, China

Upload: dyani

Post on 18-Jan-2016

28 views

Category:

Documents


0 download

DESCRIPTION

Single Image Haze Removal Using Dark Channel Prior. CVPR 2009 . Best Paper Award Kaiming He , Dept. of Inf. Eng., Chinese Univ. of Hong Kong, Hong Kong, China Jian Sun Xiaoou Tang , Dept. of Inf. Eng., Chinese Univ. of Hong Kong, Hong Kong, China. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Single Image Haze Removal Using Dark Channel Prior

Single Image Haze Removal Using Dark

Channel Prior

Professor : 王聖智 教授

Student : 戴玉書

CVPR 2009 . Best Paper Award Kaiming He, Dept. of Inf. Eng., Chinese Univ. of Hong Kong, Hong Kong, ChinaJian Sun Xiaoou Tang, Dept. of Inf. Eng., Chinese Univ. of Hong Kong, Hong Kong, China

Page 2: Single Image Haze Removal Using Dark Channel Prior

OutlineOutline

Background Background What is the Dark Channel Prior?What is the Dark Channel Prior? How to estimate How to estimate atmospheric light?? Estimating the transmission t(x) & Soft MaEstimating the transmission t(x) & Soft Ma

ttingtting Recovering the Scene RadianceRecovering the Scene Radiance ResultResult

Page 3: Single Image Haze Removal Using Dark Channel Prior

BackgroundBackground

( ) ( ) ( ) (1 ( ))I x J x t x A t x ���������������������������������������������������� ����

Observed intensity

Scene radiance

The global atmospheric light

The medium transmission,

( ) :

( ) :

:

( ) :

I x

J x

A

t x

����������������������������

( )( ) d xt x e

Page 4: Single Image Haze Removal Using Dark Channel Prior

OutlineOutline

Background Background What is the Dark Channel Prior?What is the Dark Channel Prior? How to estimate How to estimate atmospheric light? light? Estimating the transmission t(x) & Soft MaEstimating the transmission t(x) & Soft Ma

ttingtting Recovering the Scene RadianceRecovering the Scene Radiance ResultResult

Page 5: Single Image Haze Removal Using Dark Channel Prior

Dark Channel PriorDark Channel Prior Observation on haze-free outdoor images: Observation on haze-free outdoor images: In most of the non-sky patches, at least one colIn most of the non-sky patches, at least one col

or channel has very low intensity at some pixelor channel has very low intensity at some pixelss

{ , , } ( )( ) min ( min ( ( )))dark c

c r g b y xJ x J y

����������������������������

������������� �

Page 6: Single Image Haze Removal Using Dark Channel Prior

Mainly due to three factorsMainly due to three factors

ShadowsShadows Colorful of objects or surfacesColorful of objects or surfaces Dark objectsDark objects

Page 7: Single Image Haze Removal Using Dark Channel Prior

haze-free image The dark channel of haze-free image

Page 8: Single Image Haze Removal Using Dark Channel Prior
Page 9: Single Image Haze Removal Using Dark Channel Prior

Statistics of the dark Statistics of the dark channelschannels

Except for the sky region, the intensity of is low and Except for the sky region, the intensity of is low and tends to be zerotends to be zero

( )darkJ x����������������������������

Page 10: Single Image Haze Removal Using Dark Channel Prior

Visually, the intensity of the dark channel is rough

approximation of the thickness of the haze

haze image The dark channel of haze image

Page 11: Single Image Haze Removal Using Dark Channel Prior

OutlineOutline

Background Background What is the Dark Channel Prior?What is the Dark Channel Prior? To estimate of To estimate of atmospheric light light Estimating the transmission t(x) & Soft MaEstimating the transmission t(x) & Soft Ma

ttingtting Recovering the Scene RadianceRecovering the Scene Radiance ResultResult

Page 12: Single Image Haze Removal Using Dark Channel Prior

To estimate of To estimate of atmospheric light

Pick the top 0.1% brightest pixels in the dark Pick the top 0.1% brightest pixels in the dark channelchannel

Page 13: Single Image Haze Removal Using Dark Channel Prior

OutlineOutline

Background Background What is the Dark Channel Prior?What is the Dark Channel Prior? How to estimate How to estimate atmospheric light? light? Estimating the transmission t(x) & Soft MaEstimating the transmission t(x) & Soft Ma

ttingtting Recovering the Scene RadianceRecovering the Scene Radiance ResultResult

Page 14: Single Image Haze Removal Using Dark Channel Prior

Estimating the transmission Estimating the transmission

( ) ( )min ( ( )) ( ) min ( ( )) (1 ( ))c c c

y x y xI y t x J y t x A

��������������������������������������������������������

��������������������������������������������������������

( ) ( )

( ) ( )min( min ( )) ( ) min( min ( )) (1 ( ))

c c

c cc cy x y x

I y J yt x t x

A A ��������������������������������������������������������

����������������������������

( ) ( )

( ) ( )min ( ) ( ) min ( ) (1 ( ))

c c

c cy x y x

I y J yt x t x

A A ��������������������������������������������������������

����������������������������

( ) ( )

( )( ) min( min ( ( ))) min( min ( )) 0

cdark c

cc cy x y x

J yJ x J y

A ��������������������������������������������������������

��������������������������� �

( )

( )( ) 1 min( min ( ))

c

cc y x

I yt x

A ����������������������������

��������������

( )

( )( ) 1 min( min ( ))

c

cc y x

I yt x

A

����������������������������

�������������� 0.95

( ) ( ) ( ) (1 ( ))I x J x t x A t x ���������������������������������������������������� ����

Page 15: Single Image Haze Removal Using Dark Channel Prior
Page 16: Single Image Haze Removal Using Dark Channel Prior
Page 17: Single Image Haze Removal Using Dark Channel Prior

Soft MattingSoft Matting

Image matting equation:Image matting equation:

(1 )I F B �������������������������� ��

Page 18: Single Image Haze Removal Using Dark Channel Prior
Page 19: Single Image Haze Removal Using Dark Channel Prior

( ) ( ) ( )T TE t t Lt t t t t

13

|( , )

1( (1 ( ) ( ) ( )))

| | | |k

Ti jij k kk

k i j k k

I U I

410

L ij :

( )L U t t

Minimize the following cost function:

A. Levin, D. Lischinski, and Y. Weiss. A closed form solutionto natural image matting. CVPR, 1:61–68, 2006. 4, 5, 7

Page 20: Single Image Haze Removal Using Dark Channel Prior
Page 21: Single Image Haze Removal Using Dark Channel Prior
Page 22: Single Image Haze Removal Using Dark Channel Prior

OutlineOutline

Background Background What is the Dark Channel Prior?What is the Dark Channel Prior? How to estimate How to estimate atmospheric light? light? Estimating the transmission t(x) & Soft MaEstimating the transmission t(x) & Soft Ma

ttingtting Recovering the Scene RadianceRecovering the Scene Radiance ResultResult

Page 23: Single Image Haze Removal Using Dark Channel Prior

( ) ( ) ( ) (1 ( ))I x J x t x A t x ���������������������������������������������������� ����

0

( )( )

max( ( ), )

I x AJ x A

t x t

����������������������������

(t0=0.1)

Page 24: Single Image Haze Removal Using Dark Channel Prior

OutlineOutline

Background Background What is the Dark Channel Prior?What is the Dark Channel Prior? How to estimate How to estimate atmospheric light? light? Estimating the transmission t(x) & Soft MaEstimating the transmission t(x) & Soft Ma

ttingtting Recovering the Scene RadianceRecovering the Scene Radiance ResultResult

Page 25: Single Image Haze Removal Using Dark Channel Prior

ResultResult

The patch size is set to 15x15 The patch size is set to 15x15 Soft matting: Preconditioned Conjugate GSoft matting: Preconditioned Conjugate G

radient (PCG) algorithmradient (PCG) algorithm Local min operator using Marcel van HerkLocal min operator using Marcel van Herk

’s ’s fast algorithmfast algorithm

Page 26: Single Image Haze Removal Using Dark Channel Prior
Page 27: Single Image Haze Removal Using Dark Channel Prior
Page 28: Single Image Haze Removal Using Dark Channel Prior
Page 29: Single Image Haze Removal Using Dark Channel Prior
Page 30: Single Image Haze Removal Using Dark Channel Prior
Page 31: Single Image Haze Removal Using Dark Channel Prior

► Tan's resultTan's result

► Fattal's resultFattal's result

► Dark channelDark channel

Page 32: Single Image Haze Removal Using Dark Channel Prior
Page 33: Single Image Haze Removal Using Dark Channel Prior

► Tan's resultTan's result

► Fattal's resultFattal's result

► Dark channelDark channel

Page 34: Single Image Haze Removal Using Dark Channel Prior
Page 35: Single Image Haze Removal Using Dark Channel Prior

► Kopf et al's resultKopf et al's result

► Dark channelDark channel

Page 36: Single Image Haze Removal Using Dark Channel Prior

( )( ) d xt x e