ScholarBank@NUShttps://scholarbank.nus.edu.sgThe DSpace digital repository system captures, stores, indexes, preserves, and distributes digital research material.Sat, 27 Nov 2021 23:11:16 GMT2021-11-27T23:11:16Z50431- αB-spline: A linear singular blending B-splinehttps://scholarbank.nus.edu.sg/handle/10635/99468Title: αB-spline: A linear singular blending B-spline
Authors: Loe, K.F.
Abstract: A linear singular blending (LSB) technique can enhance the shape-control capability of the B-spline. This capability is derived from the blending parameters defined at the B-spline control vertices and blends LSB line segments or bilinear surface patches with the B-spline curve or surface. Varying the blending parameters between zero and unity applies tension for reshaping. The reshaped curve or surface retains the same smoothness properties as the original B-spline; it possesses the same strict parametric continuities. This is different from the β-spline, which introduces additional control to the B-spline by imposing geometrical continuities to the joints of curve segments or surface patches. For applications in which strict parametric continuities cannot be compromised, LSB provides an intuitive way to introduce tension to the B-spline.
Mon, 01 Jan 1996 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/994681996-01-01T00:00:00Z
- Uncertainty computations in neural networkshttps://scholarbank.nus.edu.sg/handle/10635/99610Title: Uncertainty computations in neural networks
Authors: Hsu, L.S.; Teh, H.H.; Chan, S.C.; Loe, K.F.
Abstract: In a three-valued neural logic network, the strengths of the nodes are confined to the ordered pairs (1,0), (0,1) and (0,0). The first two pairs represent TRUE and FALSE respectively. The meaning of the third pair depends on the type of logic used. In Kleene's logic, (0,0) represents UNKNOWN. In Bochvar's logic, it represents MEANINGLESS. In this paper we introduced neural networks that allowed the strengths to be an ordered pair of real numbers the sum of which does not exceed one. Uncertainty is expressed by having a sum of less than one. This allows us to treat uncertainties in facts, rules as well as logical operations in a unifying way.
Mon, 01 Jan 1990 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/996101990-01-01T00:00:00Z
- Two-valued neural logic networkhttps://scholarbank.nus.edu.sg/handle/10635/99609Title: Two-valued neural logic network
Authors: Hsu, L.S.; Loe, K.F.; Chan, Sing C.; Teh, H.H.
Abstract: A neural logic network that uses an ordered pair of numbers as its activation is introduced. The advantage is that it can deal with situations in which a proposition can be true, false, or unknown. The disadvantage is that computation is much more time consuming than networks whose activation consists of a single value. The network described in this paper is useful for situations where Boolean logic is sufficient.
Tue, 01 Jan 1991 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/996091991-01-01T00:00:00Z
- Temporal neural logic networkshttps://scholarbank.nus.edu.sg/handle/10635/99602Title: Temporal neural logic networks
Authors: Teh, H.H.; Hsu, L.S.; Chan, S.C.; Loe, K.F.
Abstract: Neural logic networks are generalized to cater to logical systems where the validity of rules and facts changes with time. To construct a temporal network, the validity of rules and facts is collected at a selection of time instances to determine the connecting weights of the respective instances. The weight of the temporal network is then defined as functions that would produce the known values when the proper time is substituted. Three theorems on temporal pattern recognition are proved.
Mon, 01 Jan 1990 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/996021990-01-01T00:00:00Z
- Boosting face identification in airportshttps://scholarbank.nus.edu.sg/handle/10635/78047Title: Boosting face identification in airports
Authors: Jimmy, L.J.; Loe, K.-F.
Abstract: Robust face identification system in complex airport environment, which can identify certain candidates from a crowd of people in real time, is in urgent demand. S-AdaBoost is discussed in this paper as a variant of AdaBoost to handle real world environment. The Face Identification System for Airports (F1SA), based upon S-AdaBoost algorithm, is implemented in an international airport. Comparison of results obtained from FISA with those from other leading face identification approaches based on FISA database clearly demonstrates the effectiveness of FISA in real airport environment.
Wed, 01 Jan 2003 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/780472003-01-01T00:00:00Z
- Switching hypothesized measurements: A dynamic model with applications to occlusion adaptive joint trackinghttps://scholarbank.nus.edu.sg/handle/10635/78366Title: Switching hypothesized measurements: A dynamic model with applications to occlusion adaptive joint tracking
Authors: Wang, Y.; Tan, T.; Loe, K.-F.
Abstract: This paper proposes a dynamic model supporting multimodal state space probability distributions and presents the application of the model in dealing with visual occlusions when tracking multiple objects jointly. For a set of hypotheses, multiple measurements are acquired at each time instant. The model switches among a set of hypothesized measurements during the propagation. Two computationally efficient filtering algorithms are derived for online joint tracking. Both the occlusion relationship and state of the objects are recursively estimated from the history of measurement data. The switching hypothesized measurements (SHM) model is generally applicable to describe various dynamic processes with multiple alternative measurement methods.
Wed, 01 Jan 2003 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/783662003-01-01T00:00:00Z
- An effective learning method for max-min neural networkshttps://scholarbank.nus.edu.sg/handle/10635/116689Title: An effective learning method for max-min neural networks
Authors: Teow, L.-N.; Loe, K.-F.
Abstract: Max and min operations have interesting properties that facilitate the exchange of information between the symbolic and real-valued domains. As such, neural networks that employ max-min activation functions have been a subject of interest in recent years. Since max-min functions are not strictly differentiate, we propose a mathematically sound learning method based on using Fourier convergence analysis of side-derivatives to derive a gradient descent technique for max-min error functions. This method is applied to a "typical" fuzzy-neural network model employing max-rnin activation functions. We show how this network can be trained to perform function approximation; its performance was found to be better than that of a conventional feedforward neural network.
Wed, 01 Jan 1997 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/1166891997-01-01T00:00:00Z
- Imprecise reasoning using neural networkshttps://scholarbank.nus.edu.sg/handle/10635/104574Title: Imprecise reasoning using neural networks
Authors: Hsu, Loke-Soo; Teh, Hoon-Heng; Chan, Sing-Chai; Loe, Kia Fock
Abstract: A logic is defined that weighs all available information and implements it using an emulated neural network. This allows the resulting expert system to be able to learn through examples. It also handles fuzziness in the facts and the rules, as well as the logical operations, in a natural and uniform way. It is more realistic than the certainty factor formalism which leaves out information because it takes the minimum of the certainty factors for the AND operation and maximum of the certainty factors for the OR operation. In the present scheme, all activations are weighted and taken into account. Compared with classical expert systems, the present system has the advantage of operating in two modes. In the normal mode, rules are given by experts and weights are assigned values. In the learning mode, weights are allowed to vary while the system is fed with examples.
Mon, 01 Jan 1990 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/1045741990-01-01T00:00:00Z
- Surface design via deformation of periodically swept surfaceshttps://scholarbank.nus.edu.sg/handle/10635/99421Title: Surface design via deformation of periodically swept surfaces
Authors: Tai, C.-L.; Loe, K.-F.
Abstract: In this paper sinusoidal functions are introduced to blend several contour curves to produce periodic sweep surfaces. As the basic sinusoidal functions only produce uniform periodic shapes, we introduce a unity-bounded deformation function to replace the argument of the sinusoidal functions. The deformation function provides an intuitive way to deform periodic surfaces. The technique requires only a single function to control multiple contours. The simple and intuitive control provided by the deformation function encourages users to explore alternative aesthetic designs. Variation in natural objects, such as flowers and fruits, can easily be modeled via small and random variations of the deformation function.
Mon, 01 Jan 1996 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/994211996-01-01T00:00:00Z
- Speed-up fractal image compression with a fuzzy classifierhttps://scholarbank.nus.edu.sg/handle/10635/99416Title: Speed-up fractal image compression with a fuzzy classifier
Authors: Loe, K.F.; Gu, W.G.; Phua, K.H.
Abstract: This paper presents a fractal image compression scheme incorporated with a fuzzy classifier that is optimized by a genetic algorithm. The fractal image compression scheme requires to find matching range blocks to domain blocks from all the possible division of an image into subblocks. With suitable classification of the subblocks by a fuzzy classifier we can reduce the search time for this matching process so as to speedup the encoding process in the scheme. Implementation results show that by introducing three image classes and using fuzzy classifier optimized by a genetic algorithm the encoding process can be speedup by about 40% of an unclassified encoding system. © 1997 Elsevier Science B.V.
Mon, 01 Sep 1997 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/994161997-09-01T00:00:00Z
- Multi-valued neural logic networkshttps://scholarbank.nus.edu.sg/handle/10635/99561Title: Multi-valued neural logic networks
Authors: Hsu, Loke-Soo; Teh, Hoon-Heng; Chan, Sing-Chai; Kia, Fock Loe
Abstract: Two types of networks that are useful in developing expert systems are proposed. The probabilistic network can be used for predictive types of expert systems, whereas the fuzzy network is more suitable for expert systems that help in decision making. In both cases, the expert system can operate in two modes. In the normal mode, rules are given by experts and weights are assigned values. In the learning mode, weights are allowed to vary while the system is fed with examples.
Mon, 01 Jan 1990 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/995611990-01-01T00:00:00Z
- Imprecise reasoning using neural networkshttps://scholarbank.nus.edu.sg/handle/10635/99533Title: Imprecise reasoning using neural networks
Authors: Hsu, Loke-Soo; Teh, Hoon-Heng; Chan, Sing-Chai; Loe, Kia Fock
Abstract: A logic is defined that weighs all available information and implements it using an emulated neural network. This allows the resulting expert system to be able to learn through examples. It also handles fuzziness in the facts and the rules, as well as the logical operations, in a natural and uniform way. It is more realistic than the certainty factor formalism which leaves out information because it takes the minimum of the certainty factors for the AND operation and maximum of the certainty factors for the OR operation. In the present scheme, all activations are weighted and taken into account. Compared with classical expert systems, the present system has the advantage of operating in two modes. In the normal mode, rules are given by experts and weights are assigned values. In the learning mode, weights are allowed to vary while the system is fed with examples.
Mon, 01 Jan 1990 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/995331990-01-01T00:00:00Z
- Two-valued neural logic networkhttps://scholarbank.nus.edu.sg/handle/10635/104648Title: Two-valued neural logic network
Authors: Hsu, L.S.; Loe, K.F.; Chan, Sing C.; Teh, H.H.
Abstract: A neural logic network that uses an ordered pair of numbers as its activation is introduced. The advantage is that it can deal with situations in which a proposition can be true, false, or unknown. The disadvantage is that computation is much more time consuming than networks whose activation consists of a single value. The network described in this paper is useful for situations where Boolean logic is sufficient.
Tue, 01 Jan 1991 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/1046481991-01-01T00:00:00Z
- Temporal neural logic networkshttps://scholarbank.nus.edu.sg/handle/10635/104640Title: Temporal neural logic networks
Authors: Teh, H.H.; Hsu, L.S.; Chan, S.C.; Loe, K.F.
Abstract: Neural logic networks are generalized to cater to logical systems where the validity of rules and facts changes with time. To construct a temporal network, the validity of rules and facts is collected at a selection of time instances to determine the connecting weights of the respective instances. The weight of the temporal network is then defined as functions that would produce the known values when the proper time is substituted. Three theorems on temporal pattern recognition are proved.
Mon, 01 Jan 1990 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/1046401990-01-01T00:00:00Z
- Probabilistic neural-logic networkshttps://scholarbank.nus.edu.sg/handle/10635/104611Title: Probabilistic neural-logic networks
Authors: Teh, H.H.; Chan, S.C.; Hsu, L.S.; Loe, K.F.
Abstract: Summary form only given. Recently, a novel class of networks called neural-logic networks was proposed by the authors' research group to integrate the logical reasoning capability with the concepts and techniques of the conventional neural network approach. The objective of this study is to extend the idea of the neural-logic network model to incorporate the theory of probability. This model will be able to predict the probability that the matching pattern or the possible set of solutions are correct. This gives the user an indication of the degree of reliability of the conclusion.
Sun, 01 Jan 1989 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/1046111989-01-01T00:00:00Z
- Uncertainty computations in neural networkshttps://scholarbank.nus.edu.sg/handle/10635/104649Title: Uncertainty computations in neural networks
Authors: Hsu, L.S.; Teh, H.H.; Chan, S.C.; Loe, K.F.
Abstract: In a three-valued neural logic network, the strengths of the nodes are confined to the ordered pairs (1,0), (0,1) and (0,0). The first two pairs represent TRUE and FALSE respectively. The meaning of the third pair depends on the type of logic used. In Kleene's logic, (0,0) represents UNKNOWN. In Bochvar's logic, it represents MEANINGLESS. In this paper we introduced neural networks that allowed the strengths to be an ordered pair of real numbers the sum of which does not exceed one. Uncertainty is expressed by having a sum of less than one. This allows us to treat uncertainties in facts, rules as well as logical operations in a unifying way.
Mon, 01 Jan 1990 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/1046491990-01-01T00:00:00Z
- Embedding of binary trees into hypercubeshttps://scholarbank.nus.edu.sg/handle/10635/99266Title: Embedding of binary trees into hypercubes
Authors: Bier, T.; Loe, K.-F.
Abstract: We present a mathematical model of parallel computing in a hypercubical parallel computer. This is based on embedding binary trees or forests into the n-dimensional hypercube. We consider three different models corresponding to three different computing situations. First we assume that the processing time at each level of the binary tree is arbitrary, and develop the corresponding mathematical model of an embedding of a binary tree into the hypercube. Then we assume that the processing time at each level of the binary tree is the same for all processors involved at that level, and for this we develop the mathematical model of a loop embedding of a binary tree into the hypercube. The most general case is that in which only certain neighboring levels are active. Here we assume for simplicity that only the processors corresponding to two neighboring levels are active at the same time, and correspondingly we develop the mathematical model of a level embedding of a binary tree into the hypercube to cover this case. Both loop embeddings and level embeddings allow us to use the same processor several times during the execution of a program. © 1989.
Thu, 01 Jun 1989 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/992661989-06-01T00:00:00Z
- Energy-based smoothing of datahttps://scholarbank.nus.edu.sg/handle/10635/99267Title: Energy-based smoothing of data
Authors: Loe, K.-F.
Abstract: Given a set of large noisy data, a technique which generates line segments and provides an optimal smooth connection of the line segments to form a smoothing function is proposed. In this technique data are recursively subdivided into two subsets, means of the data in each subset are computed and a line segment is defined between the means of two adjacent subsets. An energy function is defined as the sum of variances of all the subsets added with a weighted (by a weighting parameter) sum of squaring the difference of means between two adjacent subsets. Optimal connections of line segments are obtained by the simulated annealing technique. The weighting parameter is introduced in the energy function to express the relative importance of minimizing local variances of data versus the smoothness of connections of line segments. By adjusting the weighting parameter, a second optimization is obtained in the sense that the smoothing function formed by the connected line segments provides the best compromise between the closeness of data to the smoothing function and the optimal smoothness of the function. © 1993.
Wed, 01 Sep 1993 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/992671993-09-01T00:00:00Z
- Logic devices based on inductive Josephson logicshttps://scholarbank.nus.edu.sg/handle/10635/99331Title: Logic devices based on inductive Josephson logics
Authors: Loe, K.F.; Ohsawa, N.; Goto, E.
Abstract: Inductive Josephson logic (IJL) is realized by using Josephson junctions as a kind of nonlinear inductance. The operational principle of inductive Josephson logic is studied from a very simple example and generalized to multiple Josephson inductive logic devices. Concepts of forward and backward are introduced to describe the states of the operations. IJLs operate in terms of current or flux; they are suitable to be used in connection with other flux-control Josephson circuit devices such as DC SQUID or DCFP. Some circuit devices based on the principle of inductive Josephson logic are given as examples of IJL logic functions. © 1988 Plenum Publishing Corporation.
Thu, 01 Dec 1988 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/993311988-12-01T00:00:00Z
- Multi-valued neural logic networkshttps://scholarbank.nus.edu.sg/handle/10635/104590Title: Multi-valued neural logic networks
Authors: Hsu, Loke-Soo; Teh, Hoon-Heng; Chan, Sing-Chai; Kia, Fock Loe
Abstract: Two types of networks that are useful in developing expert systems are proposed. The probabilistic network can be used for predictive types of expert systems, whereas the fuzzy network is more suitable for expert systems that help in decision making. In both cases, the expert system can operate in two modes. In the normal mode, rules are given by experts and weights are assigned values. In the learning mode, weights are allowed to vary while the system is fed with examples.
Mon, 01 Jan 1990 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/1045901990-01-01T00:00:00Z
- Probabilistic neural-logic networkshttps://scholarbank.nus.edu.sg/handle/10635/99583Title: Probabilistic neural-logic networks
Authors: Teh, H.H.; Chan, S.C.; Hsu, L.S.; Loe, K.F.
Abstract: Summary form only given. Recently, a novel class of networks called neural-logic networks was proposed by the authors' research group to integrate the logical reasoning capability with the concepts and techniques of the conventional neural network approach. The objective of this study is to extend the idea of the neural-logic network model to incorporate the theory of probability. This model will be able to predict the probability that the matching pattern or the possible set of solutions are correct. This gives the user an indication of the degree of reliability of the conclusion.
Sun, 01 Jan 1989 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/995831989-01-01T00:00:00Z
- A dynamic hidden Markov random field model for foreground and shadow segmentationhttps://scholarbank.nus.edu.sg/handle/10635/40817Title: A dynamic hidden Markov random field model for foreground and shadow segmentation
Authors: Wang, Y.; Loe, K.-F.; Tan, T.; Wu, J.-K.
Abstract: This paper proposes a dynamic hidden Markov random field (DHMRF) model for foreground object and moving shadow segmentation in indoor video scenes. Given an image sequence, temporal dependencies of consecutive segmentation fields and spatial dependencies within each segmentation field are unified in the novel dynamic probabilistic model that combines the hidden Markov model (HMM) and the Markov random field (MRF). An efficient approximate filtering algorithm is derived for the DHMRF model to recursively estimate the segmentation field from the history of observed images. The foreground and shadow segmentation method integrates both intensity and edge information. Moreover, models of background, shadow, and edge information are updated adaptively for nonstationary background processes. Experimental results show that the proposed approach can accurately detect moving objects and their cast shadows even in monocular grayscale video sequences.
Mon, 01 Jan 2007 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/408172007-01-01T00:00:00Z
- Handwritten digit recognition with a novel vision model that extracts linearly separable featureshttps://scholarbank.nus.edu.sg/handle/10635/43188Title: Handwritten digit recognition with a novel vision model that extracts linearly separable features
Authors: Teow, Loo-Nin; Loe, Kia-Fock
Abstract: We use well-established results in biological vision to construct a novel vision model for handwritten digit recognition. We show empirically that the features extracted by our model are linearly separable over a large training set (MNIST). Using only a linear classifier on these features, our model is relatively simple yet outperforms other models on the same data set.
Sat, 01 Jan 2000 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/431882000-01-01T00:00:00Z
- Object-oriented language for neural network simulationhttps://scholarbank.nus.edu.sg/handle/10635/111195Title: Object-oriented language for neural network simulation
Authors: Loe, K.F.; Hsu, L.S.; Chan, S.C.; Low, H.B.
Abstract: This paper describes an object oriented language for the simulation of neural networks on supercomputers. It is based on the general framework of Parallel Distributed Processing (PDP) proposed by D.E. Rumelhart and J.L. McClelland. The purpose of the design of our object oriented language is to provide a tool for the user to describe his network connections with ease. A preprocessor will translate the description into the source code of a high level language suitable for numerical computation. The latter will be compiled and executed on a supercomputer. In this way, it will inherit the user friendliness of P3 and at the same time bypass its number crunching bottleneck.
Fri, 01 Jan 1988 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/1111951988-01-01T00:00:00Z
- 3D atlas of brain connections and functional circuitshttps://scholarbank.nus.edu.sg/handle/10635/111225Title: 3D atlas of brain connections and functional circuits
Authors: Pan, Jinghong; Nowinski, Wieslaw L.; Fock, Loe K.; Dow, Douglas E.; Chuan, Teh H.
Abstract: This work aims at the construction of an extendable brain atlas system which contains: (i) 3D models of cortical and subcortical structures along with their connections; (ii) visualization and exploration tools; and (iii) structures and connections editors. A 3D version of the Talairach- Tournoux brain atlas along with 3D Brodmann's areas are developed, co-registered, and placed in the Talairach stereotactic space. The initial built-in connections are thalamocortical ones. The structures and connections editors are provided to allow the user to add and modify cerebral structures and connections. Visualization and explorations tools are developed with four ways of exploring the brain connections model: composition, interrogation, navigation and diagnostic queries. The atlas is designed as an open system which can be extended independently in other centers according to their needs and discoveries.
Wed, 01 Jan 1997 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/1112251997-01-01T00:00:00Z
- Multi-resolution brain model in multimedia environmenthttps://scholarbank.nus.edu.sg/handle/10635/111267Title: Multi-resolution brain model in multimedia environment
Authors: Pan, Jinghong; Fock, Loe Kia; Nowinski, Wieslaw L.; Dow, Douglas E.; Huan, Teh Hung
Abstract: Toward the goal of understanding the human brain composed of billions of neurons, each with up to 1000 synapses, we have developed a framework to interactively manipulate the 3D interconnections of the brain and allow exploration of related information in a hypermedia format. Students, scientists, and medical professionals can use this system to explore the brain's interconnections, and use it as an aid in diagnosing the affected connections in a case of lost brain functions. This project focuses on 3D multi-resolution models of the brain structures, their connections, and functional circuits. This eases the exploration of the 3D models, as well as provides optimal speed of manipulation. Moreover, the 3D brain models and connections are editable by the users, allowing extensions of the atlas and exploration on the results of lost interconnections. At the same time, our extension is to manage an open system which allows a user to add more structures, connections, and other multimedia information into the existing database.
Mon, 01 Jan 1996 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/1112671996-01-01T00:00:00Z
- Joint region tracking with switching hypothesized measurementshttps://scholarbank.nus.edu.sg/handle/10635/41440Title: Joint region tracking with switching hypothesized measurements
Authors: Wang, Y.; Tan, T.; Loe, K.-F.
Abstract: This paper proposes a switching hypothesized measurements (SHM) model supporting multimodal probability distributions and presents the application of the model in handling potential variability in visual environments when tracking multiple objects jointly. For a set of occlusion hypotheses, a frame is measured once under each hypothesis, resulting in a set of measurements at each time instant. A computationally efficient SHM filter is derived for online joint region tracking. Both occlusion relationships and states of the objects are recursively estimated from the history of hypothesized measurements. The reference image is updated adoptively to deal with appearance changes of the objects. The SHM model is generally applicable to various dynamic processes with multiple alternative measurement methods.
Wed, 01 Jan 2003 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/414402003-01-01T00:00:00Z
- Video segmentation based on graphical modelshttps://scholarbank.nus.edu.sg/handle/10635/41565Title: Video segmentation based on graphical models
Authors: Wang, Y.; Tan, T.; Loe, K.-F.
Abstract: This paper proposes a unified framework for spatio-temporal segmentation of video sequences. A Bayesian network is presented to model the interactions among the motion vector field, the intensity segmentation field, and the video segmentation field. The notions of distance transformation and Markov random field are used to express spatio-temporal constraints. Given consecutive frames, an optimization method is proposed to maximize the conditional probability density of the three fields in an iterative way. Experimental results show that the approach is robust and generates spatio-temporally coherent segmentation results.
Wed, 01 Jan 2003 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/415652003-01-01T00:00:00Z
- S-AdaBoost and pattern detection in complex environmenthttps://scholarbank.nus.edu.sg/handle/10635/41820Title: S-AdaBoost and pattern detection in complex environment
Authors: Jiang, J.L.; Loe, K.-F.
Abstract: S-AdaBoost is a new variant of AdaBoost and is more effective than the conventional AdaBoost in handling outliers in pattern detection and classification in real world complex environment. Utilizing the Divide and Conquer Principle, S-AdaBoost divides the input space into a few sub-spaces and uses dedicated classifiers to classify patterns in the sub-spaces. The final classification result is the combination of the outputs of the dedicated classifiers. S-AdaBoost system is made up of an AdaBoost divider, an AdaBoost classifier, a dedicated classifier for outliers, and a non-linear combiner. In addition to presenting face detection test results in a complex airport environment, we have also conducted experiments on a number of benchmark databases to test the algorithm. The experiment results clearly show S-AdaBoost's effectiveness in pattern detection and classification.
Wed, 01 Jan 2003 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/418202003-01-01T00:00:00Z
- Knowledge-driven segmentation of the central sulcus from human brain MR imageshttps://scholarbank.nus.edu.sg/handle/10635/41841Title: Knowledge-driven segmentation of the central sulcus from human brain MR images
Authors: Zuo, W.; Hu, Q.; Aziz, A.; Loe, K.; Nowinski, W.L.
Abstract: This paper presents a knowledge-driven algorithm to identify and segment the central sulcus (CS) from human brain MR images. The dataset is reformatted along the anterior and posterior commissures (AC-PC) plane first. Then, the 3D region within the two coronal planes passing through the AC and PC is defined as the region of interest (ROI) to search for all the sulci within it. The CS is the sulcus with the largest volume within the ROI. Together with the sulci, grey matter (GM) is included for the region growing in order to deal with the partial volume effect. The GM is removed through skeletonization. Experimental results are given. © 2004 IEEE.
Thu, 01 Jan 2004 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/418412004-01-01T00:00:00Z
- A sinusoidal polynomial spline and its Bezier blended interpolanthttps://scholarbank.nus.edu.sg/handle/10635/99166Title: A sinusoidal polynomial spline and its Bezier blended interpolant
Authors: Loe, K.F.
Abstract: Functional polynomials composed of sinusoidal functions are introduced as basis functions to construct an interpolatory spline. An interpolant constructed in this way does not require solving a system of linear equations as many approaches do. However there are vanishing tangent vectors at the interpolating points. By blending with a Bezier curve using the data points as the control points, the blended curve is a proper smooth interpolant. The blending factor has the effect similar to the "tension" control of tension splines. Piecewise interpolants can be constructed in an analogous way as a connection of Bezier curve segments to achieve C1 continuity at the connecting points. Smooth interpolating surface patches can also be defined by blending sinusoidal polynomial tensor surfaces and Bezier tensor surfaces. The interpolant can very efficiently be evaluated by tabulating the sinusoidal function.
Sat, 27 Jul 1996 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/991661996-07-27T00:00:00Z
- Robust vision-based features and classification schemes for off-line handwritten digit recognitionhttps://scholarbank.nus.edu.sg/handle/10635/43111Title: Robust vision-based features and classification schemes for off-line handwritten digit recognition
Authors: Teow, L.-N.; Loe, K.-F.
Abstract: We use well-established results in biological vision to construct a model for handwritten digit recognition. We show empirically that the features extracted by our model are linearly separable over a large training set (MNIST). Using only a linear discriminant system on these features, our model is relatively simple yet outperforms other models on the same data set. In particular, the best result is obtained by applying triowise linear support vector machines with soft voting on vision-based features extracted from deslanted images. © 2002 Pattern Recognition Society. Published by Elsevier Science Ltd. All rights reserved.
Tue, 01 Jan 2002 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/431112002-01-01T00:00:00Z
- Alpha-spline: A C2 continuous spline with weights and tension controlhttps://scholarbank.nus.edu.sg/handle/10635/78004Title: Alpha-spline: A C2 continuous spline with weights and tension control
Authors: Tai, C.-L.; Loe, K.-F.
Abstract: The β-spline provides bins and tension control facilities for creating geometrically continuous curves and surfaces. Although geometric continuity is a more appropriate geometric measurement of smoothness than parametric continuity, parametric continuity is still necessary in some applications. The paper proposes a new C2 continuous spline scheme called the α-spline which provides weights and tension control. The new scheme is based on blending a sequence of singular reparametrized line segments with a piecewise NURBS curve. The idea is extended to produce α-spline surfaces. © 1999 IEEE.
Fri, 01 Jan 1999 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/780041999-01-01T00:00:00Z
- Spatiotemporal video segmentation based on graphical modelshttps://scholarbank.nus.edu.sg/handle/10635/39307Title: Spatiotemporal video segmentation based on graphical models
Authors: Wang, Y.; Loe, K.-F.; Tan, T.; Wu, J.-K.
Abstract: This paper proposes a probabilistic framework for spatiotemporal segmentation of video sequences. Motion information, boundary information from intensity segmentation, and spatial connectivity of segmentation are unified in the video segmentation process by means of graphical models. A Bayesian network is presented to model interactions among the motion vector field, the intensity segmentation field, and the video segmentation field. The notion of the Markov random field is used to encourage the formation of continuous regions. Given consecutive frames, the conditional joint probability density of the three fields is maximized in an iterative way. To effectively utilize boundary information from the intensity segmentation, distance transformation is employed in local objective functions. Experimental results show that the method is robust and generates spatiotemporally coherent segmentation results. Moreover, the proposed video segmentation approach can be viewed as the compromise of previous motion based approaches and region merging approaches. © 2005 IEEE.
Sat, 01 Jan 2005 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/393072005-01-01T00:00:00Z
- A dynamic conditional random field model for foreground and shadow segmentationhttps://scholarbank.nus.edu.sg/handle/10635/39293Title: A dynamic conditional random field model for foreground and shadow segmentation
Authors: Wang, Y.; Loe, K.-F.; Wu, J.-K.
Abstract: This paper proposes a dynamic conditional random field (DCRF) model for foreground object and moving shadow segmentation in indoor video scenes. Given an image sequence, temporal dependencies of consecutive segmentation fields and spatial dependencies within each segmentation field are unified by a dynamic probabilistic framework based on the conditional random field (CRF). An efficient approximate filtering algorithm is derived for the DCRF model to recursively estimate the segmentation field from the history of observed images. The foreground and shadow segmentation method integrates both intensity and gradient features. Moreover, models of background, shadow, and gradient information are updated adaptively for nonstationary background processes. Experimental results show that the proposed approach can accurately detect moving objects and their cast shadows even in monocular grayscale video sequences. © 2006 IEEE.
Sun, 01 Jan 2006 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/392932006-01-01T00:00:00Z
- A probabilistic approach for foreground and shadow segmentation in monocular image sequenceshttps://scholarbank.nus.edu.sg/handle/10635/39733Title: A probabilistic approach for foreground and shadow segmentation in monocular image sequences
Authors: Wang, Y.; Tan, T.; Loe, K.-F.; Wu, J.-K.
Abstract: This paper presents a novel method of foreground and shadow segmentation in monocular indoor image sequences. The models of background, edge information, and shadow are set up and adaptively updated. A Bayesian network is proposed to describe the relationships among the segmentation label, background, intensity, and edge information. A maximum a posteriori - Markov random field estimation is used to boost the spatial connectivity of segmented regions. © 2005 Pattern Recognition Society. Published by Elsevier Ltd. All rights reserved.
Sat, 01 Jan 2005 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/397332005-01-01T00:00:00Z
- Robust face detection in airportshttps://scholarbank.nus.edu.sg/handle/10635/39722Title: Robust face detection in airports
Authors: Jiang, J.L.; Loe, K.-F.; Zhang, H.J.
Abstract: Robust face detection in complex airport environment is a challenging task. The complexity in such detection systems stems from the variances in image background, view, illumination, articulation, and facial expression. This paper presents the S-AdaBoost, a new variant of AdaBoost developed for the face detection system for airport operators (FDAO). In face detection application, the contribution of the S-AdaBoost algorithm lies in its use of AdaBoost's distribution weight as a dividing tool to split up the input face space into inlier and outlier face spaces and its use of dedicated classifiers to handle the inliers and outliers in their corresponding spaces. The results of the dedicated classifiers are then nonlinearly combined. Compared with the leading face detection approaches using both the data obtained from the complex airport environment and some popular face database repositories, FDAO's experimental results clearly show its effectiveness in handling real complex environment in airports.
Thu, 01 Jan 2004 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/397222004-01-01T00:00:00Z
- Rapid and automatic detection of brain tumors in MR imageshttps://scholarbank.nus.edu.sg/handle/10635/40080Title: Rapid and automatic detection of brain tumors in MR images
Authors: Wang, Z.; Hu, Q.; Loe, K.; Aziz, A.; Nowinski, W.L.
Abstract: An algorithm to automatically detect brain tumors in MR images is presented. The key concern is speed in order to process efficiently large brain image databases and provide quick outcomes in clinical setting. The method is based on study of asymmetry of the brain. Tumors cause asymmetry of the brain, so we detect brain tumors in 3D MR images using symmetry analysis of image grey levels with respect to the midsagittal plane (MSP). The MSP, separating the brain into two hemispheres, is extracted using our previously developed algorithm. By removing the background pixels, the normalized grey level histograms are calculated for both hemispheres. The similarity between these two histograms manifests the symmetry of the brain, and it is quantified by using four symmetry measures: correlation coefficient, root mean square error, integral of absolute difference (IAD), and integral of normalized absolute difference (INAD). A quantitative analysis of brain normality based on 42 patients with tumors and 55 normals is presented. The sensitivity and specificity of IAD and INAD were 83.3% and 89.1%, and 85.7% and 83.6%, respectively. The running time for each symmetry measure for a 3D 8bit MR data was between 0.1 - 0.3 seconds on a 2.4GHz CPU PC.
Thu, 01 Jan 2004 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/400802004-01-01T00:00:00Z
- Effective learning in recurrent max-min neural networkshttps://scholarbank.nus.edu.sg/handle/10635/111170Title: Effective learning in recurrent max-min neural networks
Authors: Teow, L.-N.; Loe, K.-F.
Abstract: Max and min operations have interesting properties that facilitate the exchange of information between the symbolic and real-valued domains. As such, neural networks that employ max-rain activation functions have been a subject of interest in recent years. Since max-min functions are not strictly differentiable, we propose a mathematically sound learning method based on using Fourier convergence analysis of side-derivatives to derive a gradient descent technique for max-min error functions. We then propose a novel recurrent max-min neural network model that is trained to perform grammatical inference as an application example. Comparisons made between this model and recurrent sigmoidal neural networks show that our model not only performs better in terms of learning speed and generalization, but that its final weight configuration allows a deterministic finite automation (DFA) to be extracted in a straightforward manner. In essence, we are able to demonstrate that our proposed gradient descent technique does allow max-min neural networks to learn effectively.
Wed, 01 Apr 1998 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/1111701998-04-01T00:00:00Z
- Topographical mapping forms of objects into gaussian elastic nethttps://scholarbank.nus.edu.sg/handle/10635/99442Title: Topographical mapping forms of objects into gaussian elastic net
Authors: Loe, K.F.
Abstract: A gaussian elastic net for modeling visual processing of object forms is proposed. The mechanism of the model is based on a cognitive fact that a form of a geometrical object is characterised by borders of the object. Gaussian receptive fields are used as templates which are activated by local detection of borders. The gaussian receptive fields are topological arranged such that a gross global shape of the object is mapped into the retina. By coupling gaussian receptive fields to the nodes of an elastic net, which adaptively matches the form of an object to the nodes, the gross global shape of an object is sharpened by the net as cortical processing. To achieve this, an energy function is defined. The energy function depends on the neighbourhood distances of the nodes and the activation of the gaussian receptive fields coupling to the nodes of the net. The object form is mapped into the net and sharpened by adaptively modifying the energy function to reach an optimal value. Simulation was done with an irregular object to show the processing of the net for an arbitrary object.
Sat, 01 Jun 1996 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/994421996-06-01T00:00:00Z
- Haptic interface of a web-based training system for interventional radiology procedureshttps://scholarbank.nus.edu.sg/handle/10635/41839Title: Haptic interface of a web-based training system for interventional radiology procedures
Authors: Ma, X.; Lu, Y.; Loe, K.F.; Nowinski, W.L.
Abstract: The existing web-based medical training systems and surgical simulators can provide affordable and accessible medical training curriculum, but they seldom offer the trainee realistic and affordable haptic feedback. Therefore, they cannot offer the trainee a suitable practicing environment. In this paper, a haptic solution for interventional radiology (IR) procedures is proposed. System architecture of a web-based training system for IR procedures is briefly presented first. Then, the mechanical structure, the working principle and the application of a haptic device are discussed in detail. The haptic device works as an interface between the training environment and the trainees and is placed at the end user side. With the system, the user can be trained on the interventional radiology procedures - navigating catheters, inflating balloons, deploying coils and placing stents on the web and get surgical haptic feedback in real time.
Thu, 01 Jan 2004 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/418392004-01-01T00:00:00Z
- Haptic vascular modeling and visualization in web-enabled interventional neuroradiology simulation systemhttps://scholarbank.nus.edu.sg/handle/10635/41840Title: Haptic vascular modeling and visualization in web-enabled interventional neuroradiology simulation system
Authors: Lu, Y.; Ma, X.; Loe, K.; Chui, C.; Nowinski, W.L.
Abstract: Virtual reality system for Minimally Invasive Surgery (MIS) is a challenging problem in the context of the World Wide Web. In this paper, we present a framework of web-enabled interventional neuroradiology simulation system with force feedback. Based on the hierarchical information from segmented human vascular images, we produce the small data-size control mesh of the vasculature and finally get a smooth vascular model. When a collision occurs, we calculate the volume of force feedback according to physical parameters under which the collision occurs and give the trainee a haptic feedback by the force feedback hardware that connects to the simulation system. Our method has three features: 1) the vascular model exhibits little memory consumption; 2) the vascular model delivers good rendering performance; 3) the collision detection along with force feedback computation model is a distributed one and can provide good real time reaction to the user. The initial result obtained from applying the method in our prototype of a web-enabled simulation system is encouraging: the 3D visualization of human vasculature and the haptic feedback mechanism present the trainee a vivid surgical simulation environment and the real-time force reaction is also an exciting feature for web-enabled surgical simulation system.
Thu, 01 Jan 2004 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/418402004-01-01T00:00:00Z
- A probabilistic method for foreground and shadow segmentationhttps://scholarbank.nus.edu.sg/handle/10635/40824Title: A probabilistic method for foreground and shadow segmentation
Authors: Wang, Y.; Tan, T.; Loe, K.-F.
Abstract: This paper presents a probabilistic method for foreground segmentation that distinguishes moving objects from their cast shadows in monocular indoor image sequences. The models of background, shadow, and edge information are set up and adaptively updated. A Bayesian framework is proposed to describe the relationships among the segmentation label, background, intensity, and edge information. A Markov random field is used to boost the spatial connectivity of the segmented regions. The solution is obtained by maximizing the posterior probability density of the segmentation field.
Wed, 01 Jan 2003 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/408242003-01-01T00:00:00Z