Home | Login
Lectures       Previous announcements
Select year: 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017
Seminars in 2007
2007-12-29
    This paper was published Pattern Analysys and Machine Intelligence 2003. Result of this paper is 3D reconstruction using matched features. Abstract Many vision tasks rely upon the identification of sets of corresponding features among different images. This paper presents a method that, given some corresponding features in two stereo images, matches them with features extracted from a second stereo pair captured from a distant viewpoint. The proposed method is based on the assumption that the viewed scene contains two planar surfaces and exploits geometric constraints that are imposed by the existence of these planes to first transfer and then match image features between the two stereo pairs. The resulting scheme handles point and line features in a unified manner and is capable of successfully matching features extracted from stereo pairs that are acquired from considerably different viewpoints. Experimental results are presented, which demonstrate that the performance of the proposed method compares favorably to that of epipolar and tensor-based approaches.
2007-12-22
    I will present for gait recogniton. Title is "Gait Recognition Based on Fusion of Multi-view Gait Sequences". This paper presented ICB 2006. abstract In recent years, many gait recognition algorithms have been developed, but most of them depend on a specific view angle. In this paper,we present a new gait recognition scheme based on multi-view gait sequence fusion. An experimental comparison of the fusion of gait sequences at different views is reported. Our experiments show the fusion of gait sequences at different views can consistently achieve better results. The Dempster-Shafer fusion method is found to give a great improvement. On the other hand, we also find that fusion of gait sequences with an angle difference greater than or equal to 90◦ can achieve better improvement than fusion of those with an acute angle difference.
2007-12-15
    Abstract: A man-made environment is characterized by many parallel lines and orthogonal edges. In this article, a new method for detecting the three mutually orthogonal directions of such an environment is presented. Since real-time performance is not necessary for architectural applications, such as building reconstruction, a computationally intensive approach was chosen. However, this enables us to avoid one fundamental error of most other existing techniques. Compared to theirs, our approach is furthermore more rigorous, since all conditions given by three mutually orthogonal directions are identified and utilized. We assume a partly calibrated camera with unknown focal length and unknown principal point. By examine these camera parameters, which can be determined from orthogonal directions, falsely detected vanishing points may be rejected.
2007-12-08
    On this weekend, the title of paper for regular saturday seminar is "Putting Objects in Perspective" presented in CVPR2006. The authors are Derek Hoiem, Alexei A. Efros and Martial Hebert in Carnegie Mellon University, Robotics Institute. The author presents local object detection in the context of the overall 3D scene by modeling the interdependence of objects, surface orientations, and camera viewpoint. If you need more informatioin. please contact to me by E-mail. Sincerely Denis
2007-12-01
    The title of reference paper for regular saturday seminar is "A decision-theoretic video conference system based on gesture recognition" presented in FGR2006. The author presents hand gesture recognition based on hidden Markov model combining motion and contextual information for speaker in a conference. They also recognize the relation of the position of the hand with other objects like a computer, notes and book. The paper is attached and plz refer it.
2007-11-24
    On this weekend, I will be present a paper which tittle is " License Plate Recognition Based On Prior Knowledge " This paper was presented in Qian Gao, Xinnian Wang and Gongfu Xie IEEE International Conference on Automation and Logistics August 18 - 21, 2007, Jinan, China The focus of this paper is a new algorithm based on improved BP (back propagation) neural network for Chinese vehicle license plate recognition (LPR) is described, Proposed approach provides: -- A solution for VLP which were degraded severely -- Remarkably differs from the traditional methods is prior knowledge of license plate to the procedure of location, segmentation and recognition -- Color collocation is used to locate the license plate -- Dimensions of each character are constant, which is used to segment the character -- Layout of the Chinese VLP is an important feature is used to construct a classifier for recognizing If you need more info. Please contact me by mail or leave message on our web-board. Sincerely Kaushik Deb
2007-11-17
    Dear friends, This week I'm going to present a paper of Andras Rovid, Annamaria R. Varkonyi-Koczy , Takeshi Hashimoto, Szilveszter Balogh, Yoshifumi Shimodaira which was presented at Instrumentation and Measurement Technology Conference - IMTC 2007 Warsaw, Poland, May 1-3, 2007. Abstract - High dynamic range of illumination may cause serious distortions and other problems in viewing andfurther processing of digital images. In this paper a new tone reproduction preprocessing algorithm is introduced which may help in developing hardly or non-viewable features and content of the images. The method is based on the synthetization of multiple exposure images from which the dense part, i.e. regions having the maximum level of detail are included in the output image. The resulted high quality HDR image makes easier the information extraction and effectively supports the further processing ofthe image. If you have any questions, please contact me.
2007-11-10
    I'll present "Optical Character Recognition Program for Images of Printed Text using a Neural Network" on this saturday. This paper was presented in "ICIT 2006. IEEE International Conference on Industrial Technology 2006. ". The authors are Velappa Ganapathy and Charles C. H. Lean in Monash University Malaysia. This paper explain a simple method using a self-organizing map neural network (SOM NN) . It describes the results of training a SOM NN to perform optical character recognition on images of printed characters. If you need more information, please check the paper or contact to me. Best wishes
2007-11-03
    On this weekend, I will present "Face Recognition by Elastic Bunch Graph Matching". This paper was presented in "PAAMI(Pattern Analysis And Machine Intelligence) 1997". The authors are Laurenz Wiskott, Jean-Marc Fellous, Nobert Kruger, and Christoph von der Malsburg. This paper introduce face recognition system from single images out of a large database containing one image per person. Faces are represented by labelded graphs, based on Gabor wavelet transform. I attached this paper. If you need more information, please contact to me.
2007-10-27
    This paper was presented CVPR 2005. This paper proposed a novel matcher for straight line segments, based on their appearance and their topological layout. And finally proposed a technique to obtain epipolar geometry from mixture fo line segments. Abstract We present a new method for matching line segments between two uncalibrated wide-baseline images. Most current techniques for wide-baseline matching are based on viewpoint invariant regions. Those methods work well with highly textured scenes, but fail with poorly textured ones. We show that such scenes can be successfully matched using line segments. Moreover, since line segments and regions provide complementary information, their combined matching allows to deal with a broader range of scenes. We generate an initial set of line segment correspondences, and then iteratively increase their number by adding matches consistent with the topological structure of the current ones. Finally, a coplanar grouping stage allows to estimate the fundamental matrix even from line segments only.
2007-10-20
    IEEE Transaction on pattern analysis and machine intelligence, vol. 25, no. 10, october 2003 Sohaib Khan and Mubarak Shan Abstract?In this paper, we address the issue of tracking moving objects in an environment covered by multiple uncalibrated cameras with overlapping fields of view, typical of most surveillance setups. In such a scenario, it is essential to establish corespondence between tracks of the same object, seen in different cameras, to recover complete information about the object. We call this the problem of consistent labeling of objects when seen in multiple cameras. We employ a novel approach of finding the limits of field of view (FOV) of each camera as visible in the other cameras. We show that, if the FOV lines are known, it is possible to disambiguate between multiple possibilities for correspondence. We present a method to automatically recover these lines by observing motion in the environment. Furthermore, once these lines are initialized, the homography between the views can also be recovered. We present results on indoor and outdoor sequences containing persons and vehicles.
2007-10-06
    2004 EUSIPCO (European Signal Processing Conference) (September 6-10, 2004, Vienna, Austria) OBJECT RECOGNITION METHODS BASED ON TRANSFORMATION COVARIANT FEATURES MATAS Jri and OBDRZALEK Stepan; Abstract Methods based on distinguished regions (transformation covariant detectable patches) have achieved considerable success in a range of object recognition, retrieval and matching problems, in still images and videos. We review the state-of-the-art, describe relationship to other recognition methods, analyse their strengths and weaknesses, and present examples of successful applications. Best regards, Hoang Hon Trinh
2007-09-29
    I'll the present the paper named "Object Detection for Hierarchical Image Classification" on this reference seminar. This paper presented in Latifur Khan and Lei Wang, Mining Multimedia and Complex Data, LNAI 2797, 2003. The focus of this paper propose using an automatic scalable object boundary detection algorithm based on edge detection and region growing techniques to accurately identify all object boundaries that appear in images, and an efficient merging algorithm for joining adjacent regions which uses an adjacency graph to avoid the over-segmentation of regions. If you need more information, please refer attached paper. Sincerely, Kim, Dae-Nyeon
2007-09-22
    I'll present the paper named "Real-time Person Tracking and Pointing Gesture Recognition for Human-Robot Interaction" on this reference seminar. It was presented in The authours try to understand what human want for communicating with helper robot. For this purpose, they recognize features, like orientation of head and pointing gesture, which are useful information for robot to understand human's meaning. If you need more info., plz refer attached paper .
2007-09-15
    On this weekend, I will be present a paper which tittle is " HSI Color Model Based Lane-Marking Detection" This paper was presented in Tsung-Ying Sun, Shang-Jeng Tsai and Vincent Chan IEEE Intelligent Transportation Systems Conference, Toronto, Canada, September 2006. PP~1168 -1172 The focus of this paper is a new method using HSI color model for lane-marking detection, HSILMD, is proposed Abstract—Lane-marking detection is one of the major concerned topics in the field of driving safety and intelligent vehicle. In this paper, a new method using HSI color model for lane-marking detection, HSILMD, is proposed. In HSILMD, full color images are converted into HSI color representation, within the region of interest (ROI) aiming to detect road surface on host vehicle, the difference of intensity distribution of a row of pixels within ROI is recorded and clustered with Fuzzy c-Means algorithm. Thresholds of intensity and saturation are selected accordingly. With simple thresholds and operations, lane markings on various road scene images are detected. Results are compared with the same scheme using RGB color model and a different scheme. Robustness of this reduced computation consumption system is observed. If you need more info. Please contact me by mail or leave message on our web-board. Sincerely Kaushik Deb
2007-09-08
    I'll present "Road-Sign Detection and Recognition Based on Support Vector Machines" on this saturday. This paper was presented in "IEEE Transactions on Intelligent Transportation Systems, Vol. 8, No. 2, June 2007". The authors are "Saturnino Maldonado-Basc?, Sergio Lafuente-Arroyo, Pedro Gil-Jim?ez, Hilario G?ez-Moreno, and Fracisco L?ez-Ferreas" in Spain. This paper propose a system that have three parts of Segmetation, Shape classification, and Recognition. The system recognize traffic sign using SVM. If you want to know results of the system, visit web site : http://roadanalysis.uah.es/page_files/publication.html . If you need more information, please check the paper or contact to me. Best wishes Heechul Lim
2007-09-01
    On this weekend, I'll present "SURF: Speeded Up Robust Features". This paper was presented in "ECCV 2006" The authors are "Herbert Bay, Tinne Tuytelaars, and Luc Van Gool". This paper propose a novel scale- and rotation-invariant interest point detector and descriptor, coined SURF(Speeded Up Robust Features). This leads to a combination of novel detection, description, and matching steps. If you need more information, please check the paper or contact to me. Best wishes Chae, Hyun Uk
2007-08-25
    I'm going to present "Simultaneous Localization and Mapping: A stereo vision based approach" on this saturday. This paper was presented IROS 2006 Beijing, China. This paper presents a combination method for solving SLAM problem. The authors use KLT for feature selection and tracking. And they introduce some method which are stereo noise filter, a robust feature validation algorithm, and loop closing for reducing the uncertainty of the system. Abstract - With limited dynamic range and poor noise performance, cameras still pose considerable challenges in the application of range sensors in the context of robotic navigation, especially in the implementation of Simultaneous Localisation and Mapping (SLAM) with sparse features. This paper presents a combination of methods in solving the SLAM problem in a constricted indoor environment using small baseline stereo vision. Main contributions include a feature selection and tracking algorithm, a stereo noise filter, a robust feature validation algorithm and a multiple hypotheses adaptive window positioning method in 'closing the loop'. These methods take a novel approach in that information from the image processing and robotic navigation domains are used in tandem to augment each other. Experimental results including a real-time implementation in an office-like environment are also presented. ¡Ø I can't upload this paper. so I'm going to send by e-mail.
2007-08-18
    Abstract—Visual surveillance using multiple cameras has attracted increasing interest in recent years. Correspondence between multiple cameras is one of the most important and basic problems which visual surveillance using multiple cameras brings. In this paper, we propose a simple and robust method, based on The principal axes of people, to match people across multiple cameras. The correspondence likelihood reflecting the similarity of pairs of principal axes of people is constructed according to the relationship between ”ground-points” of people detected in each camera view and the intersections of principal axes detected in different camera views and transformed to the same view. Our method has the following desirable properties: 1) Camera calibration is not needed. 2) Accurate motion detection and segmentation are less critical due to the robustness of the principal axis-based feature to noise. 3) Based on the fused data derived from correspondence results, positions of people in each camera view can be accurately located even when the people are partially occluded in all views. The experimental results on several real video sequences from outdoor environments have demonstrated the effectiveness, efficiency, and robustness of our method.
2007-08-04
    2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'04) - Volume 2 pp. 113-120. Extraction and Integration of Window in a 3D Building Model from Ground View Images. Sung Chun Lee, University of Southern California Ram Nevatia, University of Southern California Abstract Details of the building facades are needed for high quality fly-through visualization or simulation applications. Windows form a key structure in the detailed façade reconstruction. In this paper, given calibrated facade texture (i.e. the rectified texture), we extract and reconstruct the 3D window structure of the building. We automatically extract windows (rectangles in the rectified image) using a profile projection method, which exploits the regularity of the vertical and horizontal window placement. We classify the extracted windows using 2D dimensions and image texture information. The depth of the extracted windows is automatically computed using window classification information and image line features. A single ground view image is enough to compute 3D depths of the facade windows in our approach.
2007-07-28
    On this weekend, I will be present a paper which tittle is "Using Appearance and Context for Outdoor Scene Object Classification" This paper was presented in < IEEE International Conference on Image Processing, vol. II, pp. 1218-1221, 2005> and authors are A. Bosch, X. Mu?z and J. Mart? The focus of this paper propose a probabilistic object classifier for outdoor scene analysis. First step: solving the problem of scene context generation. Second step: a stage of general segmentation provides the segmentation of unknown regions by a bottom-strategy. Finally, the last stage tries to perform a region fusion of known and unknown segmented objects. The result is both a segmentation of the image and recognition of each segment as a given object class or as an unknown segmented object. If you need more information, please check the attached paper. Sincerely Kim, Dae-Nyeon
2007-07-21
    For reference seminar on Saturday, I'll present "Detected motion classification with a double-background and a neighborhood-based difference" detecting moving object using background subtraction. In this paper, the authors update background by long-term and short-term background. As we can easily realize, short-term background is one of the blob which belong in background only short duration. Also the authors decribe they can overcome temporal illumination changes in the image sequence by background updating based on double-background. If you need more information, please check the attached paper or contact to me. Sincerely terry.
2007-07-14
    I will present seminar on this saturday. Title of paper is 'Robust Character Recognition in Low-resolution Images and Videos'. Authors are Stephan Kopf, Thomas Haenselmann, Wolfgang Effelsberg. They research character recognition in University of Mannheim, Germany. This paper is composed of 3 steps that detection of text region, segmentation of characters, classification of characters. If you want more information, contact to me please. Heechul Lim
2007-07-07
    On this weekend, I will present "Face Detection in Color Image". This paper was presented in "PAMI(Pattern Analysis And Machine Intelligence) 2002". The authors are Rein-Lien Hsu, Mohamed Abdel-Mottaleb, Anli K.Jain. In this paper, they propose a face detection algorithm for color image in the presence of varying lighting conditions as well as complex backgrounds. They use two major modules: First module is the face localization for finding candidate and second is facial feature detection for verifying detected face candidates. I attached this paper. If you have more information, contact to me please.
2007-06-23
    This paper introduces method which is detection obstacle on road using v-disparity image. Also this paper was written, road is consisted of flat and non flat, so how to detect obstacle using this method . This paper was published in Proc. IEEE Intell. Vehicle Symp. 2002. Abstract?Many roads are not totaly planar and often present hills and valleys because of environment topography. Nevertheless, the majority of existing techniques for road obstacle detection using stereovision assumes that the road is planar. This can cause several issues : imprecision as regards the real position of obstacles as well as false obstacle detection or obstacle detection failures. In order to increase the reliability of the obstacle detection process, this paper proposes an original, fast and robust method for detecting the obstacles without using the flat-earth geometry assumption; this method is able to cope with uphill and downhill gradients as well as dynamic pitching of the vehicle. Our approach is based on the construction and investigation of the v_disparity image which provides a good representation of the geometric content of the road scene. The advantage of this image is that it provides semi-global matching and is able to perform robust obstacle detection even in the case of partial occlusion or errors committed during the matching process. Furthermore, this detection is performed without any explicit extraction of coherent structures such as road edges or lane-markings in the stereo image pair. This paper begins by explaining the construction of the v_disparity image and by describing its main properties. On the basis of this image, we then describe a robust method for road obstacle detection in the context of flat and non flat road geometry, including estimation of the relative height and pitching of the stereo sensor with respect to the road surface. The longitudinal profile of the road is estimated and the objects located above the road surface are then extracted as potential obstacles; subsequently, the accurate detection of road obstacles, in particular the position of tyre-road contact points is computed in a precise manner. The whole process is performed at frame rate with a current-day PC. Our experimental findings and comparisons with the results obtained using a flat geometry hypothesis showthe benefits of our approach. Future work will be concerned with the construction of a 3D road model and the test of the system for Stop and go applications. If you have some idea or dicuss with me, whenever you can call me. Have a good luck...^^
2007-06-16
    This paper was presented in ECCV2006. Proposed algorithms - Tracking people's ground points and Segmenting blobs using overlapping views, ground-plane homography, occlusion analysis. - Robust multi-view intergartion from imperfect segmentation - Exponential increase of state space multi targets, multi views -> searching the space by several iteration of multi-view 'segmentation' - Monitoring crowded space : building entrance, store, casino, etc.
2007-06-09
    On this weekend, I will be present a paper which tittle is " A License Plate-Recognition Algorithm for Intelligent Transportation System Applications" This paper was presented in < IEEE Transactions on Intelligent Transportation Systems, Vol. 7, No. 3, september 2006> and authors are Christos Nikolaos E. Anagnostopoulos, Ioannis E. Anagnostopoulos, Vassili Loumos and Eleftherios Kayafas. The focus of this paper is a novel segmentation technique implemented for vehicle license plate (LP) identification. << Abstract >> They use an algorithm for vehicle license plate (LP) identification The algorithm on the basis of: a) Image segmentation technique (Sliding Windows) b) Connected component analysis in conjunction with a character recognition neural network The algorithm was tested with 1334 natural scene gray level vehicle images of different background & ambient illumination (Properly segmented images were 1287 over 1334 it’s near 96.5% ) The optical character recognition (OCR) is a two layer: a) Probabilistic Neural Network (PNN) with topology 108-180-36, whose performance for entire plate recognition reached 89.1% b) PNN is trained to identify alphanumeric characters from car LP based on data obtained from algorithmic image processing If you need more info. Please contact me by mail or leave message on our web-board. Sincerely Kaushik Deb
2007-06-02
    The paper was published in the IEEE Computer Society Conference on Computer Vision and Pattern Recognition 2006(CVPR'06). Authors are Yong Xu, Hui Ji and Cornelia Fermuller. Abstract: Image texture analysis has received a lot of attention in the past years. Researchers have developed many texture signatures based on texture measurements, for the purpose of uniquely characterizing the texture. Existing texture signatures, in general, are not invariant to 3D transforms such as view-point changes and non-rigid deformations of the texture surface, which is a serious limitation for many applications. In this paper, we introduce a new texture signature, called the multifractal spectrum (MFS). It provides an efficient framework combining global spatial invariance and local robust measurements. The MFS is invariant under the bi-Lipschitz map, which includes view-point changes and non-rigid deformations of the texture surface, as well as local affine illumination changes. Experiments demonstrate that the MFS captures the essential structure of textures with quite low dimension.
2007-05-26
    This weekend (2007/05/26), An interesting paper, Yan et al "PCA-SIFT A More Distinctive Representation for Local Image Descriptors" CVPR'04, Vol. 2, pp. 506~513, will be presented by H. H. Trinh. This paper proposed a method to improve the standard SIFT(developed by Lowe) to recognize objects. The paper demonstrates that the PCA based local descriptors are more distinctive, more robust to image deformations, and more compact than the standard SIFT representatioin. You are welcome to presentation! I hopes that we will have a good meeting, and discussion. (We can refer this paper at the attached file )
2007-05-19
    I'll present a reference paper which title is "Segmentation Based Environment Modeling using a Single Image". It was written by Seung Taek Ryoo and presented in ICIAR2004. This paper suggests segmentation based environment modeling method using the parallel/ perpendicular of the plane for constructing the real-time image based view-rendering. This method can easily be implemented on an environment map and makes the environment modeling easier through extracting the depth value by the image segmentation. It becomes possible to develop an environment modeling system with a full-view through this method. As shown in figures 6, 7, the 3-dimensional environment model can easily be constructed from the image segmentation method.
2007-05-12
    I'll present a reference paper which title is "Moving shadow detection with low- and Mid-level reasong". It was written by Ajay J. Joshi, Stefan Atev, Osama Masoud, and Nikolaos Papanikolopoulos and presented in ICRA2007. Their goal in this paper is to detect moving shadows from image sequence. In this paper, they use four parameters: a) edge magnitude error, b) edge gradient direction error, c) intensity ratio, d) color error. They use 3 steps to detect moving shadows from image. First, they try to seperate foreground and background from image. Then from the rereground, they search the shadow using four parameters which parameters are decribed above. Finally they recover good foreground object by blob-level reasoning. For the final step, they use shadow and foreground blobs marked by previous step. In their result, they show nicely detected shadows with foreground object from highway 1~3, intelligent room and laboratory scene.
2007-04-28
    I'll present about detection of text on road signs this Saturday. Title of paper is 'Detection of Text on Road Signs From Video' by Wen Wu, Xilin Chen, and Jie Yang. They are member of the IEEE. Wen Wu and Jie Yang are scientist at the School of Computer Science, Carnegie Mellon University, Pittsburgh, PA. Since this paper focus on detection of text on road signs, they skip the recognition part. This paper is reported on 'IEEE Transactions on intelligent transportation systems, Vol.6,No.4, December 2005'.
2007-04-21
    I introduce my next seminar. Title of this paper is "Face Localization using Face Model trained by Unsupervised Learning" and was presented "Frontiers of Computer Vision 2007(FCV 2007)". The authors are Ji-nyun CHUNG, Taemin KIM and Hyun Seung YANG in Korea Advanced Institute of Science and Technology(KAIST), Korea. In this paper, I feel interest in contents of two. They use DoG(Difference of Gaussian) function for feature extraction and unsupervised learning for face model trained as title said. If you need more information, please contact to me.
2007-04-07
    Abstract 1. A simple approach to create a good meeting room configuration 2 .Calibration of multiple cameras in the meeting room 3. Including five steps (1) Stereo camera pairs for the room configuration and requirements of the targets, the participants of the meeting (2) Applying Tsai¡¯s algorithm to calibrate each stereo camera pair (3) Using Vicon motion capture data to transfer all local coordinates system of stereo camera pair into a global coordinates system in the meeting room (4) Calibration error analysis for the current camera and meeting room configuration (5) To improve the current camera and meeting room configuration according to the error distribution 4. Repeating these steps A good meeting room configuration and parameters of all cameras for this room configuration
2007-03-31
    On this weekend, I will be present a paper name as " Estimating Velocity Fields on a Freeway from Low-Resolution Videos The authors present to develop an algorithm for traffic speed estimation and localized in space and time from video data. This paper was presented in < IEEE Transactions on Intelligent Transportation Systems, Vol. 7, No. 4, December 2006> and authors are Young Cho and John Rice. << Abstract >> An algorithm to estimate velocity fields from low resolution video recordings is presented. The algorithm does not attempt to identify and track individual vehicles, nor does it attempt to estimate derivatives of the field of pixel intensities. Rather, a frame is compressed by obtaining an intensity profile in each lane along the direction of traffic flow. The speed estimate is then computed by searching for a best matching profile in a frame at earlier and later times. It is shown that the estimate is equivalent to a weighted median of velocities of vehicles. Because the algorithm does not need high-quality images, it is directly applicable to a compressed format digital video stream, such as .mpeg, from conventional traffic video cameras. The procedure is illustrated using a 15-min-long Video Home System recording to obtain speed estimates on a 1-mi stretch of highway I-80 in Berkeley, CA. If you need more info. Please contact me by mail or leave message on our web-board. Sincerely Kaushik Deb
2007-03-24
    Author: De-Shuang Huang Journal Title: International journal of pattern recognition and artificial intelligence (Int. j. pattern recogn. artif. intell.) Publisher: World Scientific, Singapore, SINGAPOUR (1987) (Revue) The paper proposes, based on the RBFN and KNN (or PNN), a new neural network model called radial basis probabilistic neural network (RBPNN), which are not only applicable to function approximation but also to pattern classification. PS: The paper was suggested as a presentation for saturday's seminars by Dr. De-Shuang Huang, who visited our lab a week ago.
2007-03-10
    On this week, I'm going to present about "Object Class Recognition by Unsupervised Scale-Invariant Learning". This paper appears in: Computer Vision and Pattern Recognition, Volume: 2, On page(s): II-264-II-271 vol.2, 2003. The authors are R. Fergus and P. Perona in University of Oxford and A. Zisserman in California Institute of Technology. Main author of this paper was winner of CVPR 2003 Best Paper prize. This paper present a method to learn and recognize object class models. If you need more information. please contact to me by mail or leave message on our web-board. Sincerely Dennis Abstract We present a method to learn and recognize object class models from unlabeled and unsegmented cluttered scenes in a scale invariant manner. Objects are modeled as flexible constellations of parts. A probabilistic representation is used for all aspects of the object: shape, appearance, occlusion and relative scale. An entropy-based feature detector is used to select regions and their scale within the image. In learning the parameters of the scale-invariant object model are estimated. This is done using expectation-maximization in a maximum-likelihood setting. In recognition, this model is used in a Bayesian manner to classify images. The flexible nature of the model is demonstrated by excellent results over a range of datasets including geometrically constrained classes (e.g. faces, cars) and flexible objects (such as animals).
2007-03-03
    On this weekend, I'll present "A background reconstruction for dynamic scenes" based on online clustering. The author presents their algorithm can reconstruct a background where contains bi-model or multi-model distribution and it also can use to segment motion. This paper was presented in < ICIF 2006 > and authors are Mei Xiao, Chongzhao Han and Xin Kang. If you need more info. please contact to me by mail or leave message on our web-board. Sincerely Terry << Abstrct >> Based on assumption that background would not be the parts which appear in the sequence for a short time, a background reconstruction algorithm based on online clustering was proposed in this paper. Firstly, pixels intensities are classified based on online clustering. Secondly, cluster centers and appearance probabilities of each cluster are calculated. Finally, a single or multi intensities clusters with the appearance probability greater than threshold are selected as the background pixel intensity value. Simulation results show that the algorithm can represent situation where the background contains bi-model or multi-model distribution, and motion segmentation can be performed correctly. The algorithm with inexpensive computation and low memory can accommodate the real-time need.
2007-02-24
    I'll present about VioLET system on next Saturday. Paper's title is 'Video-Based Lane Estimation and Tracking for Driver Assistance: Survey, System, and Evaluation' by Joel C. McCall and Mohan M. Trivedi. They are established the Laboratory for Intelligent and Safe Automobiles(LISA) at the University of California, San Diego(UCSD). The topic is lane-marking detection at variety situation and test using a unique instrumented vehicle. This paper is reported on 'IEEE Transactions on intelligent transportation systems, Vol.7,No.1, March 2006'.
2007-02-10
    I'll present about face authentification using SIFT features on this saturday. This paper title is "On the use of SIFT features for face authentication" and was presented "Proceeding of the 2006 Conference on Computer Vision and Pattern Recognition Workshop(CVPRW'06)". The author is Manuele Bicego, Andrea Lagorio, Enrico Grosso and Massimo Tistarelli in University of Sassari, Italy. They use SIFT features, 3-different matching methods for face authentication and tested using the BANCA database and protocol.
2007-02-03
    Abstract To navigate reliably in indoor environments, a mobile robot must know where it is. Thus, reliable position estimation is a key problem in mobile robotics. We believe that probabilistic approaches are among the most promising candidates to providing a comprehensive and real-time solution to the robot localization problem. However, current methods still face considerable hurdles. In particular, the problems encountered are closely related to the type of representation used to represent probability densities over the robot¡¯s state space. Recent work on Bayesian filtering with particle-based density representations opens up a new approach for mobile robot localization, based on these principles. In this paper we introduce the Monte Carlo Localization method, where we represent the probability density involved by maintaining a set of samples that are randomly drawn from it. By using a sampling-based representation we obtain a localization method that can represent arbitrary distributions. We show experimentally that the resulting method is able to efficiently localize a mobile robot without knowledge of its starting location. It is faster, more accurate and less memory-intensive than earlier grid-based methods.
2007-01-27
    Abstract. The problem of projective reconstruction by minimization of the 2D reprojection error in multiple images is considered. Although bundle adjustment techniques can be used to minimize the 2D reprojection error, these methods being based on nonlinear optimization algorithms require a good starting point. Quasi-linear algorithms with better global convergence properties can be used to generate an initial solution before submitting it to bundle adjustment for refinement. In this paper, we propose a factorization-based method to integrate the initial search as well as the bundle adjustment into a single algorithm consisting of a sequence of weighted least-squares problems, in which a control parameter is initially set to a relaxed state to allow the search of a good initial solution, and subsequently tightened up to force the final solution to approach a minimum point of the 2D reprojection error. The proposed algorithm is guaranteed to converge. Our method readily handles images with missing points.
2007-01-13
    Next Saturday I'm going to present about HDR (High Dynamic Range) and tone mapping In computer graphics and photography, high dynamic range imaging (HDRI) is a set of techniques that allow a far greater dynamic range of exposures (i.e. a large difference between light and dark areas) than normal digital imaging techniques. The intention of HDRI is to accurately represent the wide range of intensity levels found in real scenes ranging from direct sunlight to the deepest shadows. Tone mapping is a computer graphics technique used to approximate the appearance of high dynamic range images in media with a more limited dynamic range. Print-outs, CRT or LCD monitors, and projectors all have a very limited dynamic range. Essentially, tone mapping addresses the problem of strong contrast reduction from the scene values (radiance) to the displayable range while preserving the image details and color appearance important to appreciate the original scene content. This time I will not focus on any paper nevertheless the following papers will be used in my presentation:
      1. Sumanta Pattanaik, and Hector Yee, "Adaptive Gain Control for High Dynamic Range Image Display "
      2. Géraldine Joffre, William Puech, Frédéric Comby and Jacques Joffre "High Dynamic High Dynamic Range Images Range Images from from Digital Cameras Digital Cameras Raw Raw Data"
      3. Michael D. Grossberg and Shree K. Nayar, "High Dynamic Range from Multiple Images: Which Exposures to Combine?"
2007-01-06
    Our weekly seminar of this new year is going to be started with the paper "Facial feature detection using Haar classifiers". The authors of this paper are Philip Ian Wilson and Dr. John Fernandez. This paper was published in 'Journal of Computing Sciences in Colleges' in April 2006. The idea of this paper bases on the method of Viola and Jones for accurately and rapidly detecting faces within an image. This technique can be adapted to accurately detect facial features. However, the area of the image being analyzed for a facial feature needs to be regionalized to the location with the highest probability of containing the feature. By regionalizing the detection area, false positives are eliminated and the speed of detection is increased due to the reduction of the area examined. Haar cascade classifiers is presented in this paper as main algorithm.
News | About us | Research | Lectures