MaY/juNE 2016 1541-1672/16/$33.00 © 2016 IEEE 3Published by the IEEE Computer Society

n A t u r A L L A n G u A G e P r o c e s s i n G

Cheng-Lin Liu, Chinese Academy of Sciences

Brian Lovell, University of Queensland, Australia

Dacheng Tao, University of Technology Sydney, Australia

Massimo Tistarelli, University of Sassari, Italy

This is the second part of IEEE Intelligent Systems’ special issue on

pattern recognition (see March/April 2016 for part 1). All the articles in

this installment report advances in pattern recognition for visual data (image,

video, and multimedia).

G u e s t e d i t o r s ’ i n t r o d u c t i o n

Specifi cally, in “Boosted Cross-Domain Dictionary Learning for Visual Categoriza-tion,” Fan Zhu, Ling Shao, and Yi Fang ex-tend and transfer the classical AdaBoost for cross-domain visual categorization by inte-grating an auxiliary updating mechanism, such that the auxiliary domain data under a different data distribution are adapted to the target domain at both feature repre-sentation and classifi cation levels. The pro-posed framework works in conjunction with

a learned domain-adaptive dictionary pair so that both the auxiliary domain data rep-resentations and their distribution are op-timized to match the target domain. By it-eratively updating the weak classifi ers, the categorization system allocates more credits to “similar” auxiliary domain samples, while abandoning “dissimilar” auxiliary domain samples. Experiments on multiple transfer learning scenarios demonstrate the proposed method’s outstanding performance.

Pattern Recognition, Part 2

4 www.computer.org/intelligent IEEE INTELLIGENT SYSTEMS

G u e s t e d i t o r s ’ i n t r o d u c t i o n

In “Is 2D Unlabeled Data Adequate for Facial Expression Recognition?,” Lili Tao and Bogdan J. Matuszewski present a facial expression recognition system based on the random forest technique, which uses only very simple landmark features with the view of a possible real-time implementation on low-cost portable devices. The authors performed experiments on the BU-3DFE facial expression database to provide quantitative experimental evi- dence behind more fundamental ques-tions in facial articulation analysis, namely, the relative significance of 3D information and the importance of labeled training data in supervised learning. Experimental results show the effectiveness of the described methods and demonstrate that common assumptions about facial expression recognition are debatable.

In “Face Sketch-Photo Matching Using the Local Gradient Fuzzy Pattern,” Hiranmoy Roy and Debo- tosh Bhattacharjee propose a local

fuzzy descriptor, called Local Gradient Fuzzy Pattern (LGFP), for face sketch-photo matching. LGFP is a set of membership values generated using the restricted equivalence function (REF) of the intensity (gradient value) dif-ferences between a central pixel and its neighbors in a local window. A block-wise homogeneity-based weighted fuzzy similarity is measured using an arithmetic mean aggregation of REF. The proposed face sketch-photo matching method requires neither learning procedures nor training data, and it has achieved high rank-1 recognition rates in two public databases.

In “Heartbeat Rate Measurement from Facial Video,” Mohammad A. Haque and colleagues present an effective system for measuring heartbeat rate from facial videos. The proposed method ut i l i zes a fac ia l feature point tracking method by combining a “good feature to track” and a “super- vised descent method” to overcome the limitations of currently available

measuring systems, which assume unrealistic restriction of the environ-ment . A face qual it y assessment system is also incorporated to auto- matically discard low-quality faces that occur in a realistic video sequence, to help reduce erroneous results. Experimental results on datasets collected in realistic scenarios show that the proposed system outperforms existing video-based systems for heartbeat rate measurement.

In “Driver Gaze Estimation without Use of Eye Movement,” Lex Fridman and colleagues propose a method that classifies gaze into six regions based on facial features. The method exploits the correspondence between drivers’ eye and head movements, and avoids the challenging problem of eye tracking. On a dataset of 50 drivers from an on-road study, the method achieved an average accuracy of 91.4 percent at an average decision rate of 11 Hz.

In “Combining Region-of-Interest Extraction and Image Enhancement for Nighttime Vehicle Detection,” Hulin Kuang and colleagues combine a novel region-of-interest (ROI) extraction approach that fuses vehicle l ight detection and object proposals with a nighttime image enhancement approach based on improved multiscale retinex to extract accurate ROIs and enhance images for accurate nighttime vehicle detection. The authors’ experiments on a self-developed dataset of nighttime vehicles demonstrates the effectiveness of the proposed method, score-level multifeature fusion and ROI extraction method, and the superiority of the vehicle detection method.

Finally, in “Overlaid Arrow Detection for Labeling Regions of Interest in Biomedical Images,” K.C. Santosh and colleagues present a new template-free geometric signature-based technique for detecting arrow annotations on biomedical images, a key first step

t h e A u t h o r sCheng-Lin Liu is a professor in and the director of the National Laboratory of Pattern Recognition, Institute of Automation of Chinese Academy of Sciences (CASIA). His research interests include pattern recognition, image processing, neural networks, machine learning, and especially applications to character recognition and document analysis. Liu received a PhD in pattern recognition and intelligent control from CASIA. He is a Fellow of IEEE and the IAPR. Contact him at liucl@nlpr.ia.ac.cn.

Brian Lovell is director of the Advanced Surveillance Group in the School of ITEE, University of Queensland, Australia. His interests include face recognition, biometrics, nonlinear manifold learning, and pattern recognition. Lovell is a Fellow of the IAPR, senior member of IEEE, and voting member for Australia on the governing board of the IAPR. Contact him at lovell@itee.uq.edu.au.

Dacheng Tao is a professor of computer science in the Centre for Quantum Computation & Intelligent Systems and on the Faculty of Engineering and Information Technology at the University of Technology, Sydney. His research interests span computer vision, data science, image processing, machine learning, neural networks, and video surveillance. Tao is a Fellow of IEEE, the OSA, the IAPR, the SPIE, the BCS, and the IET. Contact him at Dacheng.Tao@uts.edu.au.

Massimo Tistarelli is a professor of computer science and director of the Computer Vision Laboratory at the University of Sassari, Italy. His research interests cover biological and artificial vision (particularly in the area of recognition, 3D reconstruction, and dynamic scene analysis), pattern recognition, biometrics, visual sensors, robotic navigation, and visuo-motor coordination. Tistarelli is scientific director of the Italian Platform for Biometric Technologies, vice president of the IAPR, member of the IEEE Biometrics Professional Certification Committee, fellow member of the IAPR, and a senior member of IEEE. Contact him at tista@uniss.it.

May/june 2016 www.computer.org/intelligent 5

IEEE Computer SocietyPublications Office

10662 Los Vaqueros Circle, PO Box 3014

Los Alamitos, CA 90720-1314

Lead Editor Bonnie Wylie

b.wylie@computer.org

Editorial Management Jenny Stout

Publications Coordinator isystems@computer.org

Director, Products & Services Evan Butterfield

Senior Manager, Editorial Services Robin Baldwin

Digital Library Marketing Manager Georgann Carter

Senior Business Development Manager Sandra Brown

Senior Advertising Coordinator Marian Anderson

manderson@computer.org

Submissions: For detailed instructions and formatting, see the author guidelines at www.computer.org/intelligent/author. htm or log onto IEEE Intelligent Systems’ author center at Manuscript Central (www.computer.org/mc/intelligent/ author.htm). Visit www.computer.org/intelligent for editorial guidelines.

Editorial: Unless otherwise stated, bylined articles, as well as product and service descriptions, reflect the author’s or firm’s opinion. Inclusion in IEEE Intelligent Systems does not necessarily constitute endorsement by the IEEE or the IEEE Computer Society. All submissions are subject to editing for style, clarity, and length.

IEE

E

to ROI labeling and image content analysis. Images are first binarized using a fuzzy tool, and candidates are selected based on connected component analysis. For each candidate, arrow signatures are extracted from key points associated with its boundary and then compared with theoretical arrow signatures to determine an arrow’s presence. Experiments on the imageCLEFmed benchmark collection demonstrate the superiority of the proposed over existing methods.

We thank all the authors who submitted their invaluable

works to the special issue and all the

reviewers for their insightful comments in reviewing the submitted papers. We especially thank Daniel Zeng, the editor in chief of IEEE Intelligent Systems for giving us the opportunity of guest editing this special issue and for his continuous support. We also thank the editorial staff, Jennifer Stout, for her patience and generous help in the entire process.

Selected CS articles and columns are also available for free at http://

ComputingNow.computer.org.