automatic photo pop-up derek hoiem alexei a. efros martial hebert

21
Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

Post on 21-Dec-2015

228 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

Automatic Photo Pop-up

Derek Hoiem

Alexei A. Efros

Martial Hebert

Page 2: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

ABSTRACT Presents a fully automatic method for creating

a 3D model from a single photograph.

Algorithm labels regions of the input image into coarse categories: “ground”, “sky”, and “vertical”.

Labels are then used to “cut and fold” the image into a pop-up model using a set of simple assumptions.

The algorithm is not expected to work on every image.

Page 3: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

OUTLINEOUTLINE

INTRODUCTION RELATED WORK(暫無 ) APPROACH EXTENDING TO VIDEOS RESULTS

Page 4: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

INTRODUCTION Advances in the field of image-based rendering

during the past decade.

The method for creating virtual walkthroughs that is completely automatic and requires only a single photograph as input.

The approach is similar to the creation of a pop-up illustration in a children’s book.

Page 5: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

RELATED WORK

Page 6: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

APPROACH

Page 7: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

APPROACH

A. Handling the Illuminant Color in Images

A fundamental limitation of our previous algorithm is that it required target images to be taken under the D65.

We apply the color-by-correlation method to estimate the illuminant color.

Page 8: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

APPROACH

B. Color Naming Method We denote the categorization as the “color naming” of a pixel.

The color naming method consists of two steps:

initial color naming and fuzzy color naming. Initial color naming, for example, if the color of a pixel is categorized

in the first BCC, then the vector becomes

In the initial color naming process, each pixel completely belongs to one of the 11 BCCs.

Fuzzy color naming is done to avoid pseudo-contours.

Page 9: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert
Page 10: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

APPROACH

B. Color Naming Method

Page 11: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

APPROACH

C. Computing Corresponding Color Values in the Chromatic Categories

Page 12: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

APPROACH

D. Computing Corresponding Color Values In the Achromatic Categories

E. Transferring Colors

Page 13: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

EXTENDING TO VIDEOS

A. Color Naming Method

Page 14: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

EXTENDING TO VIDEOS

B. Computing Corresponding Color Values in the Chromatic Categories

Page 15: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

EXTENDING TO VIDEOS

C. Computing Corresponding Color Values in the Achromatic Categories

D. Transferring Colors

Page 16: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

RESULTS

Page 17: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

RESULTS

Page 18: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

RESULTS

Page 19: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

RESULTS

Page 20: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

RESULTS

Page 21: Automatic Photo Pop-up Derek Hoiem Alexei A. Efros Martial Hebert

Thank you for your listening!

2008.05.21