United States Home | Select a Location
Site Map | Contact Us | About Intel
Advanced Search
Home ComputingBusinessDeveloperReseller / Provider
Research Home
Virtual Research Center
Research On
Open Source Downloads
News & Events
Intel Technology Journal

Microprocessor Research Labs
Microprocessor Research Labs Home
Mission / Overview
Other Intel Labs

Search Intel Research

Open Source

Computer vision library debuts at CVPR

Intel (Santa Clara, CA) chose June�s Computer Vision and Pattern Recognition (CVPR) conference held at Hilton Head Island, SC to announce it�s Computer Vision Library � a set of image processing functions optimized to run on Intel microprocessors. Unlike Microsoft�s Vision software development kit that includes classes and functions for working with images, but no image-processing functions, Intel�s Computer Vision Library includes image processing, recognition, measurement, and geometric image processing functions.

Freely downloadable from (Note: This link will take you from the Intel Web site. Intel does not control the content on this Web site.) the 53.8Mbyte file also includes camera calibration, face recognition and feature tracking applications.

Started two years ago by the Intel�s Visual Interactivity Group, this is the first open source release of the Vision Library. "Our goal is to establish an open source vision community and provide a site where the distributed efforts of the community can be consolidated and performance optimized," says Gary R. Bradski of Intel's Microprocessor Research Laboratory.

"To involve and benefit from the combined expertise of the vision community," says Bradski, "we have recruited a committee of experts to participate in acceptance decisions for new content for the library and in setting the library direction," he says. Intel's contribution is to provide research, host the library on its website, and contribute assembly language optimized versions of the most compute intensive code.

Two of the applications using the image processing functions contained within the library are automatic camera calibration and face recognition. Developed by Gary Bradski, Jean-Yves Bouguet and Vadim Pisarevsky, the camera calibration tool automatically tracks a calibration object and then applies a calibration algorithm developed by Zhengyou Zhang at Microsoft Research (Redmond, WA). Developed by Ara Nefian and Monson Hayes at Georgia Tech Lorraine (Metz, France), theThe Vision Library�s camera calibration tool allows any video camera to be calibrated in a few seconds.

"In operation, " says NefianBouguet, "A flat checkerboard pattern is placed in front of the camera, and the program automatically acquires a number of images. These are then used to compute focal length, principal point, and distortion coefficients of the camera and the three-dimensional position of the pattern for each image. "Since the corners of the pattern are located automatically on each image, the entire procedure is automatic," says NefianBouguet. "Once calibration is complete, the program allows video images to be undistorted in real timeat frame rates.

Using an algorithm developed by Ara Nefian and Monson Hayes at Georgia Tech (Atlanta, GA)Zhengyou Zhang at Microsoft Research (Redmond, WA), the Vision Library�s face recognition tool lets Intel-based PCs read a face from BMP image files or from an USB-based camera. "Previous approaches to face recognition include geometric feature-based methods, template-based methods, and more recently model-based methods, our approach uses a statistical network," says ZhangNefian.

"Significant facial features include hair, forehead, eyes, nose and mouth - features that occur in a natural order, from left to right, top to bottom, even if the images undergo small rotations," he Ara says. Because of this, facial images can be modeled using a pseudo twoone-dimensional statistical model known as an embedded Hidden Markov Model (eHMM) by assigning each of these regions to a state in a 2D grid. In such model, the states themselves are not observable, but rather yield observation vectors that are statistically dependent on the state of the eHMM. These vectors are used in the face recognition process. According to Nefian, this one-dimensionalembedded HMM achieves recognition rates of about 85%98% to 100% on the ORL face database.

FIGURE CAPTION: Applications within Intel�s freely available Computer Vision Library include a camera calibration tool that allows parameters such as focal length, principal point, and distortion coefficients to be computed.





Back to Top

*Legal Information  |  Privacy Policy
©2002 Intel Corporation