Convert 2D Images to 3D Image
No Thumbnail Available
Date
2013
Journal Title
Journal ISSN
Volume Title
Publisher
Uva Wellassa University of Sri Lanka
Abstract
Image processing plays a major role in modern technology. Implementations of this field are
varying from simple applications like documentation to complex applications like space
context. Before use in the above applications images should be processed using relevant
techniques. There are many software that can process the images and get the outputs to be used
in the above mentioned applications. Conversion of 2D images into 3D format is essential in
applications such as automatic navigation of robots and vehicles, satellite identification and
fault diagnosis, medical reasoning and remote surgery (Pollefeys et al., 2000). With developed
software user can obtain 3D images using 2D images in a lesser time and the human interaction
have been eliminated to a single step.
Methodology
MATLAB is a technical computing language used throughout the research (Background on
matlab image processing toolboxes, 2013). The system allows user to input two stereo images
of the object which is to be converted into 3D. Then the image pair is read and converted to
gray scale for the matching process. To have more efficient and accurate results two system
objects were used.
Basically red-cyan channel images give 3D visualization when visualized with 3D spectacles
or one with a red and cyan vision. So the conversion of two stereo images to cyan and red is
done. RGB value of cyan is 0,255,255 and RGB value of red is 255, 0, 0. So left image is
(:,:,1:1)=0 and right image is(:,:,2:3)=0;.
For the best block matching step, extract the 7 X 7 pixel block is considered in the right image
and search along the same row in the left image (Image point matching, 2013). Sum of absolute
difference is used to compare the image regions. Template Matcher System object is used to
increase the accuracy of blocks matching (www.imageprocessingbook.com). Then allocating
space for all template matcher system objects is done. Scan is done over all rows and columns
and disparity bound is calculated for each. Then template and region of interest is constructed.
Depth map have positive values only. Finally As there are noisy patches and bad depth
estimates everywhere noise filtration is done.
Description
Keywords
Science and Technology, Technology, Robotics, Software Developing, Computer Science, Modern Technology.