Researcher applies optical correlator to 3-D scenes

EE Times Home
This Week's News

Search

Site Archive
Print Archive
HotBot

Career Corner

J.O.B.S. Online
Ask the Headhunter
Salary Survey

Columns

Wolfe's Den
Core Competency
Weaving your Web
Europhile
Institutional Memory

Features

Smart Technologies
Special Reports

Fun & Games

Immortal Works
Game Power

News Center
Technology
Business & Finance
Products
All Headlines

EE Design Center
EE Design Online
Reference Library
PLD Center
Prototype Center
IP Yellow Pages

Career Center
Jobs!

EDTN Direct
EDTN on Pointcast

Network Publications
EE Times
EBN
SBN
Product File

Home

Headlines

Careers

Columns

IP Watch

Gimme more

Subscribe to our print edition. Conference coverage of recent industry shows. Future conferences. Contact our editors.

Posted: 5:30 p.m., EDT, 8/31/98

Researcher applies optical correlator to 3-D scenes

Sunny Bains

STANFORD, Calif. — In a bid to solve some of the fundamental problems in image recognition, a scientist in Israel has extended optical correlator techniques from two dimensions to three.

The scheme, which involves fusing images of the object from many points of view, makes it possible to locate objects in 3-D space, removing some of the ambiguity inherent in the current generation of 2-D optical correlators. In particular, the device is not confused by objects of different sizes that seem identical only because they are positioned at different distances from the observer.

The inventor is Joseph Rosen, senior lecturer in electrical and computer engineering at Ben-Gurion University of the Negev. Rosen said the technique will make object tracking easier in general.

"Suppose objects in the observed scene are moving in a complicated path to the left (or right) and toward (or away from) the observer," said Rosen. "The 3-D correlation will enable us to analyze this true exact track. Moreover, if some objects, or their details, are covered by other objects there is a chance that covered information will be exposed when the scene is observed from different points of view."

The new approach impresses Joseph Goodman, Stanford University EE professor, whose book Fourier Optics is considered the definitive text in the field. "Optical signal processing is by now relatively mature," said Goodman. "In a field that is nearly 40 years old, significant new concepts arise relatively infrequently. Rosen's work on 3-D correlation is in this category."

Optical correlators determine whether two images are similar or identical by comparing their Fourier transforms. Transforms are generated optically, by presenting a collimated laser image to a lens. Though such correlators are proving useful in their two-dimensional form, the fact that they are scale- and orientation-dependent has made them less than ideal for some applications. Though the new system still has the problem of orientation dependence, the scaling problem becomes less of an issue in the 3-D case.

The correlator developed at Ben-Gurion University uses a standard charge-coupled device (CCD) electronic camera, or a series of them, to take pictures of both the scene to be examined and the reference object. If there are several cameras, each is aimed at the same point in the 3-D scene from a different direction. With one camera, a series of images is taken sequentially, with the position and point of view changing from one image to the next.

The basic design uses parallel optical Fourier transforms that can be performed with a lens. However, in the experiments soon to be described in an upcoming issue of the journal Applied Optics, the FTs were done digitally, because Rosen didn't have a spatial light modulator to code the incoming images onto a light beam.

The algorithm, said Rosen, "yields a direct 3-D correlation of realistic 3-D scenes without the need first to reconstruct and to understand the scene." In addition, using optics to perform the two-dimensional Fourier transforms will give the technique some speed advantages. "The optical 2-D-FT is done in parallel," he said, "by a device as simple as a lens and at the speed of light."

The next stage involves getting a 3-D transform out of the two-dimensional ones. The most obvious way to get a 3-D Fourier transform would involve first taking "slice" images through the depth of the scene in question and performing a 2-D FT on each of those. After that, a one-dimensional FT would be carried out on the set of corresponding pixels from each image—for example, the top right pixel from each depth slice. When every "depth line" has been transformed in this way, the 3-D transform is finished. The last step is necessary in this case because no depth information is available via a single 2-D slice.

The situation in the system demonstrated by Rosen, however, is very different. Clearly the cameras cannot capture 2-D slices of the scene. Instead the images captured are projections from the scene at slightly different angles. Among them, these images contain information about all three dimensions; they simply have to be placed into the correct context in 3-D space. This is done using a coordinate transformation, performed electronically.

Doing the FT is just the first half of correlating a reference image with a target scene. After the two have been jointly transformed, they have to be transformed back to produce a correlation peak.

Though Rosen has successfully demonstrated this technique using an entirely digital correlator, he believes the system can be improved by using optics. In his new scheme, 2-D FTs would be done optically at both ends of the system, with the coordinate transformation and a 1-D FT for the third dimension performed electronically. His next step will be to build such a hybrid system to demonstrate the concept.

"I seriously think that there are no real technical challenges to implementing the proposed system," Rosen said. "I just need a moderate budget to purchase a few CCDs and spatial light modulators."

Home

Headlines

Careers

Columns

IP Watch

EE Times Services

EE Times Online provides a number of helpful services for our readers and advertisers. The EE Times Network is a portfolio that includes an editorial calendar, a media kit and a list of trade shows and conferences.

is a vehicle to request information from vendors mentioned in recent EE Times print ads and articles.
Product Shopper J.O.B.S. Online Search Firm Directory Media Kit Marketplace

Register

Please become a registered user of EDTN. It is free and will make it easier for you to receive the information you need. See the benefits of registration, or:

Register Now

All material on this site Copyright © 1998 CMP Media Inc. All rights reserved.

Search EDTN and the Top 100 Electronics Sites


EE Times Home This Week's News Search Site Archive Print Archive HotBot Career Corner J.O.B.S. Online Ask the Headhunter Salary Survey Columns Wolfe's Den Core Competency Weaving your Web Europhile Institutional Memory Features Smart Technologies Special Reports Fun & Games Immortal Works Game Power News Center Technology Business & Finance Products All Headlines EE Design Center EE Design Online Reference Library PLD Center Prototype Center IP Yellow Pages Career Center Jobs! EDTN Direct EDTN on Pointcast Network Publications EE Times EBN SBN Product File	Home	Headlines	Careers	Columns	IP Watch
						Gimme more
						Subscribe to our print edition. Conference coverage of recent industry shows. Future conferences. Contact our editors.
	Posted: 5:30 p.m., EDT, 8/31/98 Researcher applies optical correlator to 3-D scenes Sunny Bains STANFORD, Calif. — In a bid to solve some of the fundamental problems in image recognition, a scientist in Israel has extended optical correlator techniques from two dimensions to three. The scheme, which involves fusing images of the object from many points of view, makes it possible to locate objects in 3-D space, removing some of the ambiguity inherent in the current generation of 2-D optical correlators. In particular, the device is not confused by objects of different sizes that seem identical only because they are positioned at different distances from the observer. The inventor is Joseph Rosen, senior lecturer in electrical and computer engineering at Ben-Gurion University of the Negev. Rosen said the technique will make object tracking easier in general. "Suppose objects in the observed scene are moving in a complicated path to the left (or right) and toward (or away from) the observer," said Rosen. "The 3-D correlation will enable us to analyze this true exact track. Moreover, if some objects, or their details, are covered by other objects there is a chance that covered information will be exposed when the scene is observed from different points of view." The new approach impresses Joseph Goodman, Stanford University EE professor, whose book Fourier Optics is considered the definitive text in the field. "Optical signal processing is by now relatively mature," said Goodman. "In a field that is nearly 40 years old, significant new concepts arise relatively infrequently. Rosen's work on 3-D correlation is in this category." Optical correlators determine whether two images are similar or identical by comparing their Fourier transforms. Transforms are generated optically, by presenting a collimated laser image to a lens. Though such correlators are proving useful in their two-dimensional form, the fact that they are scale- and orientation-dependent has made them less than ideal for some applications. Though the new system still has the problem of orientation dependence, the scaling problem becomes less of an issue in the 3-D case. The correlator developed at Ben-Gurion University uses a standard charge-coupled device (CCD) electronic camera, or a series of them, to take pictures of both the scene to be examined and the reference object. If there are several cameras, each is aimed at the same point in the 3-D scene from a different direction. With one camera, a series of images is taken sequentially, with the position and point of view changing from one image to the next. The basic design uses parallel optical Fourier transforms that can be performed with a lens. However, in the experiments soon to be described in an upcoming issue of the journal Applied Optics, the FTs were done digitally, because Rosen didn't have a spatial light modulator to code the incoming images onto a light beam. The algorithm, said Rosen, "yields a direct 3-D correlation of realistic 3-D scenes without the need first to reconstruct and to understand the scene." In addition, using optics to perform the two-dimensional Fourier transforms will give the technique some speed advantages. "The optical 2-D-FT is done in parallel," he said, "by a device as simple as a lens and at the speed of light." The next stage involves getting a 3-D transform out of the two-dimensional ones. The most obvious way to get a 3-D Fourier transform would involve first taking "slice" images through the depth of the scene in question and performing a 2-D FT on each of those. After that, a one-dimensional FT would be carried out on the set of corresponding pixels from each image—for example, the top right pixel from each depth slice. When every "depth line" has been transformed in this way, the 3-D transform is finished. The last step is necessary in this case because no depth information is available via a single 2-D slice. The situation in the system demonstrated by Rosen, however, is very different. Clearly the cameras cannot capture 2-D slices of the scene. Instead the images captured are projections from the scene at slightly different angles. Among them, these images contain information about all three dimensions; they simply have to be placed into the correct context in 3-D space. This is done using a coordinate transformation, performed electronically. Doing the FT is just the first half of correlating a reference image with a target scene. After the two have been jointly transformed, they have to be transformed back to produce a correlation peak. Though Rosen has successfully demonstrated this technique using an entirely digital correlator, he believes the system can be improved by using optics. In his new scheme, 2-D FTs would be done optically at both ends of the system, with the coordinate transformation and a 1-D FT for the third dimension performed electronically. His next step will be to build such a hybrid system to demonstrate the concept. "I seriously think that there are no real technical challenges to implementing the proposed system," Rosen said. "I just need a moderate budget to purchase a few CCDs and spatial light modulators."
	Home	Headlines	Careers	Columns	IP Watch	EE Times Services
						EE Times Online provides a number of helpful services for our readers and advertisers. The EE Times Network is a portfolio that includes an editorial calendar, a media kit and a list of trade shows and conferences. is a vehicle to request information from vendors mentioned in recent EE Times print ads and articles. Product Shopper J.O.B.S. Online Search Firm Directory Media Kit Marketplace
						Register
						Please become a registered user of EDTN. It is free and will make it easier for you to receive the information you need. See the benefits of registration, or: Register Now
	All material on this site Copyright © 1998 CMP Media Inc. All rights reserved.					Search EDTN and the Top 100 Electronics Sites