Depth Edge Detection Using Multiflash Imaging
A report by Sheira Ben Haim and Dvir Segal,
April 2014.
This project is based on the paper: "Non-photorealistic
Camera: Depth Edge and Stylized Rendering using Multi-Flash Imaging" [Ramesh
et al. 2004]
w Introduction to the problem
In this project we deal with the problem of directly
finding depth edges or create stylized images based on depth discontinuities.
The purpose is being able to convey 3D
structure of imaged scene using only 2D camera capabilities.
w Our Approach
We present a rendering approach to
capture and convey shape features of real-world scenes. We use a model that
mimics a camera with multiple flashes.
The use of multiple flashes helps us in casting shadows along depth
discontinuities in the scene. The projective-geometric relationship of the
camera-flash setup is then exploited to detect depth discontinuities and
distinguish them from intensity edges due to material discontinuities.
In addition, we implement a method for
utilizing the detected edge features to generate stylized images. The resulting
images more clearly convey the 3D structure of the imaged scenes.
This approach of capturing geometric
features is different than traditional approaches that require reconstructing a
3D model.
w Algorithm
Creating depth edges image:
Input: 4 images, each captured with a
different flash position.
Preparations:
1.
Make gray scale
images out of the input images.
2.
Normalize the
intensity for all gray scale images, represented by
.
Calculate:
3.
Compute the maximum
intensity image 
.
4.
For each,
compute ratio image
,
such that 
.
5.
For each image, find
the pixels with step edges with negative transition, and mark them as depth
edges pixels.
6.
Create the Depth
Edge Image – by creating an image combined of all four images in last step,
taking the maximum value found for each pixel
7.
Create 4, 2-level
canny-like hysteresis images, and combine them to obtain the final binary
edge image.
Outputs:
1.
Depth Edge Image
2.
Binary Edge Image
Stylizing the image:
Input: the Binary Edge
Image
A.
Emphasized image
1.
Use image
morphology in order to improve the edges image:
a.
Create a square
structure
b.
Dilate the edges of
the input image according to the structure
c.
Ask the user to
mark regions in order to fill in holes.
d.
Erode back the
filled image according to the structure.
2.
Create an
Emphasized image using the last step output.
B.
Styled image
1.
Use the input image
to find out connected components.
2.
Average the color
of each component – to eliminate texture.
Outputs:
1.
An Emphasized image
– based on one of the four input images emphasized by depth edges.
2.
A Styled Image - a
synthesized image created by abstracting the texture.
w Implementation
1.
In order to find
the depth edges we are using the Sobel kernel, generated by fspecial, to
find the derivative of the intensity, applying it on each by imfilter.
2.
For creating the 4
edge images , we use a threshold method implemented by bwlabel in
the following manner:
a)
Find low and high
contours which are higher than the given confidence value.
b)
Compute all
connected components of both images using bwlabel.
3.
In order to enhance
and improve and edged image, in the emphasizing part, we are using the strel
function, Creating a morphological structuring element.
We are using the structure for dilating
and eroding
the edged image while filling in
areas of holes.
w Results\Examples
Loaded images
Flash Down Flash Left
Flash Right Flash
Up
Results
Edge detected Binary Image
Emphasized image Stylized
image
Additional Results
Edge detected Binary Image
Emphasized image Stylized
image
Edge detected Binary Image
Emphasized image Stylized
image
Edge detected Binary Image
Emphasized image Stylized
image
w Who wrote this and why?
This Project submitted by Sheira Ben
Haim and Dvir Segal in Computational Photography course as part of their B.Sc. in
Computer Science at the University of Haifa.
The project was performed under the
supervision of Hagit Hel-Or.
w Terms of use
Copyright © 2014 by Sheira Ben Haim
and Dvir Segal.
The SOFTWARE ("Depth Edge
Detection Using Multiflash Imaging") is provided "as is",
without any guarantee made as to its suitability or fitness for any particular
use. It may contain bugs, so use of this tool is at your own risk. We take no
responsibility for any damage that may unintentionally be caused through its
use.
You may not use the source to SOFTWARE
or distribute SOFTWARE in any form, without express written permission of the
copyright holders.