Gal A. Kaminka's Publications

Sorted by DateClassified by Publication TypeClassified by TopicGrouped by Student (current)Grouped by Former Students

Perception of a 3D Colored Image from One Colored and One Gray-Scale Images

Yael Termin. Perception of a 3D Colored Image from One Colored and One Gray-Scale Images. Ph.D. Thesis, Bar Ilan University, 2007.

Download

(unavailable)

Abstract

Stereoscopic vision is one of the most interesting tasks used in daily life. It facilitates depth perception used in order to safely navigate in our three--dimensional (3D) world to self-orient in space, and to visually recognize objects. Color vision supplies us with important additional information. The percept generated by the human visual system is far more complex than the sum of both images seen by the two eyes. It is well known that depth perception can be obtained by viewing a stereo pair that was acquired from slightly different angles of view through a stereoscope. A stereo pair consisting of two monochromatic (gray-scale) images yields a monochromatic 3D image and a stereo pair consisting of two color images fused into a colorful 3D image. Computer vision systems and remote controlled teleoperation systems use algorithms that combines color and depth. Both features are important for obstacle and target detection and benefits are gained from combining both. When both eyes are presented with dissimilar stimuli the visual system is "confused" and does not always manage to fuse them into a single stable percept. This unstable phenomenon is called binocular rivalry. Previous works have involved images with different features being presented to the two eyes: Contrast, different content (house vs. face), different line direction and two different colors. However, a number of questions were left open: What if one of the color images within a color stereo pair is replaced by a gray-scale image? Will depth perception remain intact even though chromatic data is absent from one of the stereo images pairs? Will color perception still be valid? This thesis addresses these open questions. Controlled psychophysical experiments validated the hypothesis that a stereo pair consisting of one gray-scale image and one color image (will be referred to as MIX) produces 3D color perception. Four basic combinations were presented to each subject: 1) Full color stereo pair (color image on both sides); 2) Full monochromatic pair (gray-scale image on both sides); 3) Two MIX stereo pair (color image on the right side and a gray-scale image on the left side and vice versa). Various image sets were presented to the subjects with two apparatus: A stereoscope an HMD (Head Mounted Display). All subjects perceived a colored image with depth. No degradation in depth perception was measured, although some degradation in the perception of color was observed. It is already known that the two images in a stereo pair are not required to have equal sharpness in order to perceive depth. A certain amount of blur on one of the images within a gray-scale pair or color pair will not affect the depth perception obtained from both images. We have found that when the gray-scale image is blurred, depth perception remains nearly the same while color perception improves. An effect of color perception asymmetry while viewing the MIX stereo pair was found. The interposition of the color image within the stereo pair (right or left image) influenced the color perception. Most subjects reported that when color image was presented to the right eye, color perception was better then when the color image was presented to the left eye. We further investigated adding rivalrous patterns of diagonal lines (inclined +45 degrees on one image and -45 degrees on the other image) to both images of the MIX pair. While the direction of the line was altered, depth and color perception was stable. Additional experiments were conducted with various compression techniques applied to the gray-scale image. Results have shown that the basic effect can endure compression of various sorts. Utilizing the findings reported in this study will allow the perception of color and depth while reducing the number of channels needed to be transmitted or stored from six channels to only four channels. In addition, compression if added to the gray-scale image can further decrease the amount of data. The research presented explores the minimal requirements necessary for viewing a stereoscopic color image, from a psychophysical aspect. These finding might be relevant to solving various computer vision tasks like bandwidth reduction and an efficient storage of color stereoscopic images.

Additional Information

BibTeX

@PhdThesis{termin-phd, 
  author = 	 {Yael Termin}, 
  title = 	 {Perception of a 3D Colored Image from One Colored and One Gray-Scale Images}, 
  school = 	 {{B}ar {I}lan {U}niversity}, 
  year = 	 {2007}, 
  OPTkey = 	 {}, 
  OPTtype = 	 {}, 
  OPTaddress = 	 {}, 
  OPTmonth = 	 {}, 
  OPTnote = 	 {}, 
  abstract = { 
Stereoscopic vision is one of the most interesting tasks used in 
daily life. It facilitates depth perception used in order to 
safely navigate in our three--dimensional (3D) world to 
self-orient in space, and to visually recognize objects. Color 
vision supplies us with important additional information. The 
percept generated by the human visual system is far more complex 
than the sum of both images seen by the two eyes. 
It is well known that depth perception can be obtained by viewing 
a stereo pair that was acquired from slightly different angles of 
view through a stereoscope. A stereo pair consisting of two 
monochromatic (gray-scale) images yields a monochromatic 3D image 
and a stereo pair consisting of two color images fused into a 
colorful 3D image. Computer vision systems and remote controlled teleoperation 
systems use algorithms that combines color and depth. Both 
features are important for obstacle and target detection and 
benefits are gained from combining both. 
When both eyes are presented with dissimilar stimuli the visual 
system is "confused" and does not always manage to fuse them into 
a single stable percept. This unstable phenomenon is called  
binocular rivalry. Previous works have involved images with 
different features being presented to the two eyes: Contrast, 
different content (house vs. face), different line direction and 
two different colors. 
However, a number of questions were left open: What if one of the 
color images within a color stereo pair is replaced by a 
gray-scale image? Will depth perception remain intact even though 
chromatic data is absent from one of the stereo images pairs? Will 
color perception still be valid? This thesis addresses these open 
questions. 
Controlled psychophysical experiments validated the hypothesis 
that a stereo pair consisting  of one gray-scale image and one 
color image (will be referred to as MIX) produces 3D color 
perception. Four basic combinations were presented to each 
subject: 1) Full color stereo pair (color image on both sides); 2) 
Full monochromatic pair (gray-scale image on both sides); 3) Two 
MIX stereo pair (color image on the right side and a gray-scale 
image on the left side and vice versa). Various image sets were 
presented to the subjects with two apparatus: A stereoscope an HMD 
(Head Mounted Display). All subjects perceived a colored image 
with depth. No degradation in depth perception was measured, 
although some degradation in the perception of color was observed. 
It is already known that the two images in a stereo pair are not 
required to have equal sharpness in order to perceive depth. A 
certain amount of blur on one of the images within a gray-scale 
pair or color pair will not affect the depth perception obtained 
from both images. We have found that when the gray-scale image is 
blurred, depth perception remains nearly the same while color 
perception improves. 
An effect of color perception asymmetry while viewing the MIX 
stereo pair was found. The interposition of the color image within 
the stereo pair (right or left image) influenced the color 
perception. Most subjects reported that when color image was 
presented to the right eye, color perception was better then when 
the color image was presented to the left eye. 
We further investigated adding rivalrous patterns of diagonal 
lines (inclined +45 degrees on one image and -45 degrees on the 
other image) to both images of the MIX pair. While the direction 
of the line was altered, depth and color perception was stable. 
Additional experiments were conducted with various compression   
techniques applied to the gray-scale image. Results have shown 
that the basic effect can endure compression of various sorts. 
Utilizing the findings reported in this study will allow the 
perception of color and depth while reducing the number of 
channels needed to be transmitted or stored from six channels to 
only four channels. In addition, compression if added to the 
gray-scale image can further decrease the amount of data. The 
research presented explores the minimal requirements necessary for 
viewing a stereoscopic color image, from a psychophysical aspect. 
These finding might be relevant to solving various computer vision 
tasks like bandwidth reduction and an efficient storage of color 
stereoscopic images.}, 
  wwwnote = {}, 
  OPTannote = 	 {} 
} 

Generated by bib2html.pl (written by Patrick Riley ) on Sat Feb 24, 2018 00:31:02