Yuan-Kai Wang | FuJen University (original) (raw)

Papers by Yuan-Kai Wang

2019 IEEE 8th Global Conference on Consumer Electronics (GCCE)

Nail painting machine is a new kind of consumer device getting prevailing, but a smart technique ... more Nail painting machine is a new kind of consumer device getting prevailing, but a smart technique to automatically find nail printing area becomes necessary. In this study, we propose an image segmentation approach to mark the area of nails, which cuts the nail image and merges the painting patterns selected by the user. The segmented-and merged result is then sent to the nail painting machine for printing.

2013 International Conference on Machine Learning and Cybernetics, 2013

Automatic sleep pattern analysis has been a very important research issue for the diagnosis in sl... more Automatic sleep pattern analysis has been a very important research issue for the diagnosis in sleep medicine. This paper proposes a nonintrusive sleep/wake identification method based on computer vision approach to extract visual sleep activity and sleep/wake patterns. This approach is robust to noise, contrast and illumination variations of infrared videos. The proposed method extracts body motion context by illumination compensation and background subtraction algorithms, and sleep status is recognized by linear regression of body motion context. Experiments are conducted on the video polysomnography data from 18 persons recorded in sleep laboratory. The sleep/wake status identified from the infrared videos is verified with the ground truth that is scored by a sleep technician from the polysomnography data according to standard medical operation. High accuracy of the experiments demonstrates the validity of the proposed method.

IEEE Transactions on Image Processing, 2014

IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2017

2016 International Conference on Machine Learning and Cybernetics (ICMLC), 2016

Abstract Business card recognition is a difficult problem. Characters in business card are small ... more Abstract Business card recognition is a difficult problem. Characters in business card are small with diverse font types. An approach using the left-right hidden Markov model is proposed for business card recognition. The hidden Markov model will output a top-10 ...

Abstract Illumination and expression variations are still a challenging problem in face recogniti... more Abstract Illumination and expression variations are still a challenging problem in face recognition. In this work, we present an efficient face recognition method which can solve the above two problems with single training sample. At first, the effect of the lighting variation ...

Abstract Character normalization recovers distortions occurring in character images. There are tw... more Abstract Character normalization recovers distortions occurring in character images. There are two kinds of character distortions: local and global. Local distortion has been discussed in the literature, but the global distortion that is usually produced by geometric ...

Abstract In this paper, genetic algorithms are applied to solve the error-correcting subgraph iso... more Abstract In this paper, genetic algorithms are applied to solve the error-correcting subgraph isomorphism (ECSI) problems. The error-correcting subgraph isomorphism problems are first formulated as permutation searching problems. Two ECSI algorithms are devised. ...

Page 1. Applying Genetic Algorithms on Pattern Recognition: An Analysis and SurveyYuan-Kai Wang&a... more Page 1. Applying Genetic Algorithms on Pattern Recognition: An Analysis and SurveyYuan-Kai Wang' and Kuo-Chin Fan' Institute of Information Science, Academia Sinica, Taipei, Taiwan, ROC Institute of Computer Science ...

Pattern Recognition, Nov 1, 2002

... skew documents. It usually appears in the front of a large and dark region around the margin ... more ... skew documents. It usually appears in the front of a large and dark region around the margin of document images. Marginal noise might cover meaningful document objects, such as text, graphics and forms. The overlapping ...

Signal Processing, Sep 1, 1995

Document analysis plays an important role in office automation, especially in intelligent signal ... more Document analysis plays an important role in office automation, especially in intelligent signal processing. In this paper, we propose an intelligent document analysis system to achieve the document segmentation and identification goal. The proposed system ...

Pattern Recognition, Nov 1, 2000

Pattern Recognition, Dec 1, 1997

Kanerva's Sparse Distributed Memory (SDM) is one of the self-organizing neural netwo... more Kanerva's Sparse Distributed Memory (SDM) is one of the self-organizing neural networks that mimic closely the psychological behavior of the human brain. In this paper, a Genetic Sparse Distributed Memory (GSDM) model that combines SDM with genetic algorithms is ...

Applied sciences, May 16, 2019

Textile Research Journal, Mar 1, 1998

Fabric classification plays an important role in the textile industry. In this paper, two fabric ... more Fabric classification plays an important role in the textile industry. In this paper, two fabric classification methods, the neural network and dimensionality reduction, are proposed to automatically classify fabrics based on measured hand properties. The methods are independent and reinforce each other. The first method adopts a neural network to recognize the category of an unknown fabric. In the second method, a dimensionality reduction technique is applied to reduce the dimensionality of the mea sured properties of input fabrics from sixteen dimensions to two. The reduced features are then plotted in a two-dimensional coordinate system to visualize and verify the classification results of the neural network. In experiments conducted to verify the validity of our proposed approach, fabric data are expressed in the form of hand prop erties extracted from the KES-FB system (Kawabata's evaluation system for fabrics). These experiments confirm the feasibility and efficiency of our approach with a wide variety of fabrics.

Page 1. A Fuzzy Bipartite Weighted Graph Matching Appro;rch to Fingerprint C'erit?c:... more Page 1. A Fuzzy Bipartite Weighted Graph Matching Appro;rch to Fingerprint C'erit?c:ttion Kuo-Chin Fan Cheng-Wen Liu Y uan-Kai W;ing Institute of Coniputer Science Institute of Computer Science Institute of Iiiformation Science, ...

Abstract In this paper, we present a method based on Active Appearance Model (AAM) to remove eyeg... more Abstract In this paper, we present a method based on Active Appearance Model (AAM) to remove eyeglasses from face images. The occluded regions are first roughly detected by the AAM search. Then, an ellipse model is used to fit the eyes' position. After eliminating the ...

Journal of Real-Time Image Processing, 2012

ABSTRACT Retinex is an image restoration approach used to restore the original appearance of an i... more ABSTRACT Retinex is an image restoration approach used to restore the original appearance of an image. Among various methods, a center/surround retinex algorithm is favorable for parallelization because it uses the convolution operations with large-scale sizes to achieve dynamic range compression and color/lightness rendition. This paper presents a GPURetinex algorithm, which is a data parallel algorithm accelerating a modified center/surround retinex with GPGPU/CUDA. The GPURetinex algorithm exploits the massively parallel threading and heterogeneous memory hierarchy of a GPGPU to improve efficiency. Two challenging problems, irregular memory access and block size for data partition, are analyzed mathematically. The proposed mathematical models help optimally choose memory spaces and block sizes for maximal parallelization performance. The mathematical analyses are applied to three parallelization issues existing in the retinex problem: block-wise, pixel-wise, and serial operations. The experimental results conducted on GT200 GPU and CUDA 3.2 showed that the GPURetinex can gain 74 times acceleration, compared with an SSE-optimized single-threaded implementation on Core2 Duo for the images with 4,096 × 4,096 resolution. The proposed method also outperforms the parallel retinex implemented with the nVidia Performance Primitives library. Our experimental results indicate that careful design of memory access and multithreading patterns for CUDA devices should acquire great performance acceleration for real-time processing of image restoration.

2014 IEEE International Conference on Consumer Electronics - Taiwan, 2014

High dynamic range image processing have recently become an important topic in consumer electroni... more High dynamic range image processing have recently become an important topic in consumer electronics market. While multi-scale retinex with color restoration (MSRCR) have been well developed, disadvantages of low performance is not favorable to a mobile computer-vision system. To remedy the above problem, this paper proposes an accelerated MSRCR with effective use of ARM Cortex-A9 architecture and NEON SIMD technology. A linear sampling method with binomial normal approximation is developed for improving performance of Gaussian smoothing. Overall performance improvement of MSRCR algorithm on Zedboard platform is 74% compared to original ARM optimized C code compiled to Cortex-A9 processor architecture.

2019 IEEE 8th Global Conference on Consumer Electronics (GCCE)

Nail painting machine is a new kind of consumer device getting prevailing, but a smart technique ... more Nail painting machine is a new kind of consumer device getting prevailing, but a smart technique to automatically find nail printing area becomes necessary. In this study, we propose an image segmentation approach to mark the area of nails, which cuts the nail image and merges the painting patterns selected by the user. The segmented-and merged result is then sent to the nail painting machine for printing.

2013 International Conference on Machine Learning and Cybernetics, 2013

Automatic sleep pattern analysis has been a very important research issue for the diagnosis in sl... more Automatic sleep pattern analysis has been a very important research issue for the diagnosis in sleep medicine. This paper proposes a nonintrusive sleep/wake identification method based on computer vision approach to extract visual sleep activity and sleep/wake patterns. This approach is robust to noise, contrast and illumination variations of infrared videos. The proposed method extracts body motion context by illumination compensation and background subtraction algorithms, and sleep status is recognized by linear regression of body motion context. Experiments are conducted on the video polysomnography data from 18 persons recorded in sleep laboratory. The sleep/wake status identified from the infrared videos is verified with the ground truth that is scored by a sleep technician from the polysomnography data according to standard medical operation. High accuracy of the experiments demonstrates the validity of the proposed method.

IEEE Transactions on Image Processing, 2014

IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2017

2016 International Conference on Machine Learning and Cybernetics (ICMLC), 2016

Abstract Business card recognition is a difficult problem. Characters in business card are small ... more Abstract Business card recognition is a difficult problem. Characters in business card are small with diverse font types. An approach using the left-right hidden Markov model is proposed for business card recognition. The hidden Markov model will output a top-10 ...

Abstract Illumination and expression variations are still a challenging problem in face recogniti... more Abstract Illumination and expression variations are still a challenging problem in face recognition. In this work, we present an efficient face recognition method which can solve the above two problems with single training sample. At first, the effect of the lighting variation ...

Abstract Character normalization recovers distortions occurring in character images. There are tw... more Abstract Character normalization recovers distortions occurring in character images. There are two kinds of character distortions: local and global. Local distortion has been discussed in the literature, but the global distortion that is usually produced by geometric ...

Abstract In this paper, genetic algorithms are applied to solve the error-correcting subgraph iso... more Abstract In this paper, genetic algorithms are applied to solve the error-correcting subgraph isomorphism (ECSI) problems. The error-correcting subgraph isomorphism problems are first formulated as permutation searching problems. Two ECSI algorithms are devised. ...

Page 1. Applying Genetic Algorithms on Pattern Recognition: An Analysis and SurveyYuan-Kai Wang&a... more Page 1. Applying Genetic Algorithms on Pattern Recognition: An Analysis and SurveyYuan-Kai Wang' and Kuo-Chin Fan' Institute of Information Science, Academia Sinica, Taipei, Taiwan, ROC Institute of Computer Science ...

Pattern Recognition, Nov 1, 2002

... skew documents. It usually appears in the front of a large and dark region around the margin ... more ... skew documents. It usually appears in the front of a large and dark region around the margin of document images. Marginal noise might cover meaningful document objects, such as text, graphics and forms. The overlapping ...

Signal Processing, Sep 1, 1995

Document analysis plays an important role in office automation, especially in intelligent signal ... more Document analysis plays an important role in office automation, especially in intelligent signal processing. In this paper, we propose an intelligent document analysis system to achieve the document segmentation and identification goal. The proposed system ...

Pattern Recognition, Nov 1, 2000

Pattern Recognition, Dec 1, 1997

Kanerva's Sparse Distributed Memory (SDM) is one of the self-organizing neural netwo... more Kanerva's Sparse Distributed Memory (SDM) is one of the self-organizing neural networks that mimic closely the psychological behavior of the human brain. In this paper, a Genetic Sparse Distributed Memory (GSDM) model that combines SDM with genetic algorithms is ...

Applied sciences, May 16, 2019

Textile Research Journal, Mar 1, 1998

Fabric classification plays an important role in the textile industry. In this paper, two fabric ... more Fabric classification plays an important role in the textile industry. In this paper, two fabric classification methods, the neural network and dimensionality reduction, are proposed to automatically classify fabrics based on measured hand properties. The methods are independent and reinforce each other. The first method adopts a neural network to recognize the category of an unknown fabric. In the second method, a dimensionality reduction technique is applied to reduce the dimensionality of the mea sured properties of input fabrics from sixteen dimensions to two. The reduced features are then plotted in a two-dimensional coordinate system to visualize and verify the classification results of the neural network. In experiments conducted to verify the validity of our proposed approach, fabric data are expressed in the form of hand prop erties extracted from the KES-FB system (Kawabata's evaluation system for fabrics). These experiments confirm the feasibility and efficiency of our approach with a wide variety of fabrics.

Page 1. A Fuzzy Bipartite Weighted Graph Matching Appro;rch to Fingerprint C'erit?c:... more Page 1. A Fuzzy Bipartite Weighted Graph Matching Appro;rch to Fingerprint C'erit?c:ttion Kuo-Chin Fan Cheng-Wen Liu Y uan-Kai W;ing Institute of Coniputer Science Institute of Computer Science Institute of Iiiformation Science, ...

Abstract In this paper, we present a method based on Active Appearance Model (AAM) to remove eyeg... more Abstract In this paper, we present a method based on Active Appearance Model (AAM) to remove eyeglasses from face images. The occluded regions are first roughly detected by the AAM search. Then, an ellipse model is used to fit the eyes' position. After eliminating the ...

Journal of Real-Time Image Processing, 2012

ABSTRACT Retinex is an image restoration approach used to restore the original appearance of an i... more ABSTRACT Retinex is an image restoration approach used to restore the original appearance of an image. Among various methods, a center/surround retinex algorithm is favorable for parallelization because it uses the convolution operations with large-scale sizes to achieve dynamic range compression and color/lightness rendition. This paper presents a GPURetinex algorithm, which is a data parallel algorithm accelerating a modified center/surround retinex with GPGPU/CUDA. The GPURetinex algorithm exploits the massively parallel threading and heterogeneous memory hierarchy of a GPGPU to improve efficiency. Two challenging problems, irregular memory access and block size for data partition, are analyzed mathematically. The proposed mathematical models help optimally choose memory spaces and block sizes for maximal parallelization performance. The mathematical analyses are applied to three parallelization issues existing in the retinex problem: block-wise, pixel-wise, and serial operations. The experimental results conducted on GT200 GPU and CUDA 3.2 showed that the GPURetinex can gain 74 times acceleration, compared with an SSE-optimized single-threaded implementation on Core2 Duo for the images with 4,096 × 4,096 resolution. The proposed method also outperforms the parallel retinex implemented with the nVidia Performance Primitives library. Our experimental results indicate that careful design of memory access and multithreading patterns for CUDA devices should acquire great performance acceleration for real-time processing of image restoration.

2014 IEEE International Conference on Consumer Electronics - Taiwan, 2014

High dynamic range image processing have recently become an important topic in consumer electroni... more High dynamic range image processing have recently become an important topic in consumer electronics market. While multi-scale retinex with color restoration (MSRCR) have been well developed, disadvantages of low performance is not favorable to a mobile computer-vision system. To remedy the above problem, this paper proposes an accelerated MSRCR with effective use of ARM Cortex-A9 architecture and NEON SIMD technology. A linear sampling method with binomial normal approximation is developed for improving performance of Gaussian smoothing. Overall performance improvement of MSRCR algorithm on Zedboard platform is 74% compared to original ARM optimized C code compiled to Cortex-A9 processor architecture.

本演講將探討影像辨識技術與雲端運算技術兩者的結合與應用。智慧型視訊監控(Intelligent Video Surveillance, IVS)為近年來相當重要的兩項應用，與雲端結合後衍生出VS... more 本演講將探討影像辨識技術與雲端運算技術兩者的結合與應用。智慧型視訊監控(Intelligent Video Surveillance, IVS)為近年來相當重要的兩項應用，與雲端結合後衍生出VSAAS(Video Surveillance As A Service)的概念與營運模式。本演講將說明智慧型視訊監控技術、應用、與相關發展，說明與雲端結合之VSAAS的觀念，並闡明如何以GPGPU將影像辨識進行平行化以達到高效能運算的目的。

Markerless human pose estimation is an important method that provides non-intrusive and high-free... more Markerless human pose estimation is an important method that provides non-intrusive and high-free motion capture for the development of behavior recognition. This talk presents a novel human motion capture method that locates human body joint position and reconstructs the human pose in 3D space from monocular images. It is a two-stage framework including 2D and 3D probabilistic graphical models which can solve the occlusion problem for the estimation of human joint positions. The 2D and 3D models adopt directed acyclic structure to avoid error propagation of inference in the models. Image observations corresponding to shape and appearance features of humans are considered as evidence for the inference of 2D joint positions in the 2D model. The estimation result of the 2D model in the first stage is regarded as the observations of the 3D model in the second stage to improve the estimation result of 3D joint positions. Both the 2D and 3D models utilize the Expectation Maximization algorithm to learn prior distributions of the models. An annealed Gibbs sampling method is proposed for the two-stage method to inference the maximum posteriori distributions of joint positions. The annealing process can efficiently explore the mode of distributions and find solutions in high-dimensional space. Experiments are conducted on the HumanEva dataset to show the effectiveness of the proposed method. The experimental data are image sequences of walking motion with a full 180 turn around a region, which causes occlusion of poses and loss of image observations. Experimental results show that the presented two-stage approach can efficiently estimate more accurate human poses.

Embedded computer vision is a greatly interesting research topic rising in recent years. It is th... more Embedded computer vision is a greatly interesting research topic rising in recent years. It is the research of combing embedded systems and computer vision for distributed computation. Smart camera, vision sensor network, and robotic vision are few of application examples of embedded computer vision. Some leading research groups in the world, such as Stanford Univ., CMU, and Princeton, have been devoted to study the hardware and software architectures for embedded computer vision.
This talk will give an introduction to the concept of embedded computer vision. Recent works of the Intelligent System Lab., EE, FJU, in the ECV research topic will be presented. The presented results include a gesture recognition system called X-EYE, a small vision sensor node called FJUCam, face detection and recognition for robotic vision, and FPGA implementation of visual surveillance algorithms.