Multi-scale local binary pattern histograms for face recognition
4 stars based on
Information and Computing Sciences: Research in biometrics carried out at Surrey since has generated IP relating to a number of aspects of automatic face recognition, which resulted in significant td investment trading account improvement, rendering this biometric technology commercially exploitable.
Through an IP agreement, these innovations have been commercially exploited multiscale local binary pattern histograms for face recognition the University spinout company OmniPerception, which has developed products for various security applications. Research in face biometrics carried out at Surrey and led by Professor Kittler since has generated Intellectual Property relating to a number of aspects of the automatic face recognition process, which resulted in significant performance improvement in face recognition, rendering this biometric technology commercially exploitable.
The advances made at Surrey include multispectral imaging and photometric normalisation to achieve illumination invariance , robust face detection and localisation using robust correlation [3,4], innovative representation of face skin texture using a multiscale local binary descriptor , a unique patented person specific discriminant analysis, which is exceptionally compact allowing face matching on a smart card , facial component based matching enhancing robustness to face localisation errors ,, and multi-algorithmic fusion, exploiting a patented error correcting decision making approach .
The innovative method of face image representation and matching  invented by Professor Kittler inwhich is now protected by a European patent, has unique properties. Its low computational complexity opened, for the first time, the possibility of implementing face verification systems on a small computing platform, such as a smart card.
This contrasts with previous solutions with computational complexity of orders of magnitude greater. In parallel, a multi-classifier system based on the concept of error correcting coding was developed at Surrey by Kittler, Ghaderi and Windeatt, for face recognition scenarios where computing power was not a constraining factor. The work was published  after filing for protection in . It has the capacity to enhance face recognition performance by a factor of two.
The intellectual property encompassed by the two patents was transferred to a university spin out company, OmniPerception Ltd. In addition, applied research carried out at Surrey after the spin out inresulted in significant enhancements of face detection  and face localisation  methods.
The former is achieved using a novel correlation method, which is robust to outliers image degradation. The latter has been developed to perform face localisation in general conditions where the pose of the face image deviates from the frontal. Another key contribution was the work on skin texture representation, carried out at Surrey in The proposed multiscale generalisation enhanced the performance of face recognition significantly.
Face verification using error correcting output codes. Multiscale local binary pattern histograms for face recognition and Vision Computing, The promise of the Intellectual Property in face biometrics generated by Professor Kittler in the late 's, and filed for protection by Surrey at the turn of the millennium, was instrumental in setting up a unique commercialization framework.
The spinout company was initially assisted both, financially, from the University seed fund and Cascade fund, and intellectually, on a long-term basis, by the University committing any pipeline IP, multiscale local binary pattern histograms for face recognition be generated by Professor Kittler after OmniPerception had been formed, for the benefit of the company.
The relationship between the University and the company mirrored the business and exploitation model used by the Stanford University in setting up the multiscale local binary pattern histograms for face recognition recognition company Nuance at about the same time.
The software engineering and product development have been facilitated by the use of common open source image processing library RAVL. This ensured that common classes and structures were used for algorithm development. In April the company merged with Visimetrics, a major UK biometrics technology integrator, to create an enterprise with the combined capability to manufacture high technology products, and to integrate them in advanced security applications.
A number of University of Surrey PhD graduates have joined the company. In April OmniPerception Ltd merged with Visimetrics, an integrator, to create a company of greater critical mass, with multiscale local binary pattern histograms for face recognition access to security markets.
With the proven track record of its security product installation at the Heathrow airport, the most recent successes of the company include the introduction of the OmniPerception face access control systems to Manchester and other UK airports by Menzies. Facial capture and search engine for police custody suits to facilitate law enforcement http: Suspect identification UK Police Forces http: Access control to air side in airports for personnel handling air cargo e.
Secure access to data centres in financial institutions http: Submitting Institution University of Surrey. Summary Impact Multiscale local binary pattern histograms for face recognition Technological.