• Journal of Computing Science and Engineering
• /
• v.9 no.4
• /
• pp.204-213
• /
• 2015

Under the conditional independence assumption among local features, the Naive Bayes Nearest Neighbor (NBNN) classifier has been recently proposed and performs classification without any training or quantization phases. While the original NBNN shows high classification accuracy without adopting an explicit training phase, the conditional independence among local features is against the compositionality of objects indicating that different, but related parts of an object appear together. As a result, the assumption of the conditional independence weakens the accuracy of classification techniques based on NBNN. In this work, we look into this issue, and propose a novel Bayesian network for an NBNN based classification to consider the conditional dependence among features. To achieve our goal, we extract a high-level feature and its corresponding, multiple low-level features for each image patch. We then represent them based on a simple, two-level layered Bayesian network, and design its classification function considering our Bayesian network. To achieve low memory requirement and fast query-time performance, we further optimize our representation and classification function, named relation-based Bayesian network, by considering and representing the relationship between a high-level feature and its low-level features into a compact relation vector, whose dimensionality is the same as the number of low-level features, e.g., four elements in our tests. We have demonstrated the benefits of our method over the original NBNN and its recent improvement, and local NBNN in two different benchmarks. Our method shows improved accuracy, up to 27% against the tested methods. This high accuracy is mainly due to consideration of the conditional dependences between high-level and its corresponding low-level features.

• The Transactions of the Korea Information Processing Society
• /
• v.4 no.12
• /
• pp.3211-3218
• /
• 1997

This paper addresses an eye location algorithm which is essential process of human face tracking system for natural video-conferencing. In current video-conferencing systems, user's facial movements are restricted by fixed camera, therefore it is inconvenient to users. We Propose an eye location algorithm for automatic face tracking. Because, locations of other facial features guessed from locations of eye and scale of face in the image can be calculated using inter-ocular distance. Most previous feature extraction methods for face recognition system are approached under assumption that approximative face region or location of each facial feature is known. The proposed algorithm in this paper uses no prior information on the given image. It is not sensitive to backgrounds and lighting conditions. The proposed algorithm uses the valley representation as major information to locate eyes. The experiments have been performed for 213 frames of 17 people and show very encouraging results.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.41 no.5
• /
• pp.13-22
• /
• 2004

Because video sequence consists of dynamic objects in nature, the object motion in video is an effective feature in describing the contents of video sequence and motion feature plays an important role in video retrieval. In this paper, we propose a method that converts motion vectors (MVs) to a uniform set on MPEG coded domain, independent of the frame type and the direction of prediction, and utilizes these normalized MVs (N-MVs) as motion descriptor to understand video contents. We describe a frame-type independent representation of the various types of frames presented in an MPEG video in which all frames can be considered equivalently, without full-decoding. In the experiments, we show that the proposed method is better than the conventional one in terms of performance.

• Journal of Institute of Control, Robotics and Systems
• /
• v.10 no.10
• /
• pp.855-868
• /
• 2004

There are several types of singularities in controlling robotic manipulators: kinematic singularity, algorithmic singularity, semi-kinematic singularity, semi-algorithmic singularity, and representation singularity. The kinematic and algorithmic singularities have been investigated intensively because they are not predictable or difficult to avoid. The problem with these singularities is an unnecessary performance reduction in non-singular region and the difficulty in performance tuning. Tn this paper, we propose a method of avoiding kinematic and algorithmic singularities by applying a task reconstruction approach while maximizing the task performance by calculating singularity measures. The proposed method is implemented by removing the component approaching the singularity calculated by using singularity measure in real time. The outstanding feature of the proposed task reconstruction method (TR-method) is that it is based on a local task reconstruction as opposed to the local joint reconstruction of many other approaches. And, this method has dynamic task priority assignment feature which ensures the system stability under singular regions owing to the change of task priority. The TR-method enables us to increase the task controller gain to improve the task performance whereas this increase can destabilize the system for the conventional algorithms in real experiments. In addition, the physical meaning of tuning parameters is very straightforward. Hence, we can maximize task performance even near the singular region while simultaneously obtaining the singularity-free motion. The advantage of the proposed method is experimentally tested by using the 7-dof spatial manipulator, and the result shows that the new method improves the performance several times over the existing algorithms.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.47 no.1
• /
• pp.1-14
• /
• 2010

With the rapid growth of 3D content industry fields, 3D content-based hashing (or hash function) has been required to apply to authentication, trust and retrieval of 3D content. A content hash can be a random variable for compact representation of content. But 3D content-based hashing has been not researched yet, compared with 2D content-based hashing such as image and video. This paper develops a robust 3D content-based hashing based on key-dependent 3D surface feature. The proposed hashing uses the block surface coefficient using shape coordinate of 3D SSD and curvedness for 3D surface feature and generates a binary hash by a permutation key and a random key. Experimental results verified that the proposed hashing has the robustness against geometry and topology attacks and has the uniqueness of hash in each model and key.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.47 no.1
• /
• pp.44-52
• /
• 2010

In this paper, we propose a novel method for face recognition which uses polar coordinate instead of conventional cartesian coordinate. Among the central area of a face, we select a point as a pole and make a polar image of a face by evenly sampling pixels in each direction of 360 degrees around the pole. By applying conventional feature extraction methods to the polar image, the recognition rates are improved. The polar coordinate delineates near-pole area more vividly than the area far from the pole. In a face, important regions such as eyes, nose and mouth are concentrated on the central part of a face. Therefore, the polar coordinate of a face image can achieve more vivid representation of important facial regions compared to the conventional cartesian coordinate. The proposed polar coordinate transform was applied to Yale and FRGC databases and LDA and NLDA were used to extract features afterwards. The experimental results show that the proposed method performs better than the conventional cartesian images.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.10
• /
• pp.5179-5196
• /
• 2019

To explore an effective non-invasion medical imaging diagnostics approach for hepatocellular carcinoma (HCC), we propose a method based on adopting the multiple technologies with the multi-parametric data fusion, transfer learning, and multi-scale deep feature extraction. Firstly, to make full use of complementary and enhancing the contribution of different modalities viz. multi-parametric MRI images in the lesion diagnosis, we propose a data-level fusion strategy. Secondly, based on the fusion data as the input, the multi-scale residual neural network with SPP (Spatial Pyramid Pooling) is utilized for the discriminative feature representation learning. Thirdly, to mitigate the impact of the lack of training samples, we do the pre-training of the proposed multi-scale residual neural network model on the natural image dataset and the fine-tuning with the chosen multi-parametric MRI images as complementary data. The comparative experiment results on the dataset from the clinical cases show that our proposed approach by employing the multiple strategies achieves the highest accuracy of 0.847±0.023 in the classification problem on the HCC differentiation. In the problem of discriminating the HCC lesion from the non-tumor area, we achieve a good performance with accuracy, sensitivity, specificity and AUC (area under the ROC curve) being 0.981±0.002, 0.981±0.002, 0.991±0.007 and 0.999±0.0008, respectively.

• Journal of Internet Computing and Services
• /
• v.20 no.6
• /
• pp.1-10
• /
• 2019

Object-based image matching algorithms have been widely used in the image processing and computer vision fields. A variety of applications based on image matching algorithms have been recently developed for object recognition, 3D modeling, video tracking, and biomedical informatics. One prominent example of image matching features is the Scale Invariant Feature Transform (SIFT) scheme. However many applications using the SIFT algorithm have implemented based on stand-alone basis, not client-server architecture. In this paper, We initially implemented based on client-server structure by using SIFT algorithms to identify and match objects in biomedical images to provide useful information to the user based on the recently released Mobile platform. The major methodological contribution of this work is leveraging the convenient user interface and ubiquitous Internet connection on Mobile device for interactive delineation, segmentation, representation, matching and retrieval of biomedical images. With these technologies, our paper showcased examples of performing reliable image matching from different views of an object in the applications of semantic image search for biomedical informatics.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.45 no.5
• /
• pp.32-37
• /
• 2008

To ensure access to rapidly growing video collection, video indexing is becoming more and more important. In this paper, video ontology system for retrieving a video data based on a scene unit is proposed. The proposed system creates a semantic scene as a basic unit of video retrieval, and limits a domain of retrieval through a subject of that scene. The content of semantic scene is defined using the relationship between object and event included in the key frame of shots. The semantic gap between the low level feature and the high level feature is solved through the scene ontology to ensure the semantic-based retrieval.

• Journal of KIISE:Software and Applications
• /
• v.31 no.12
• /
• pp.1649-1659
• /
• 2004

Text Generation is a process of generating comprehensible texts in human languages from some underlying non-linguistic representation of information. Among several sub-processes for text generation to generate coherent texts, this paper concerns referring expression generation which produces different types of expressions to refer to previously-mentioned things in a discourse. Specifically, we focus on pronominalization by zero pronouns which frequently occur in Korean. To build a generation model of referring expressions for Korean, several features are identified based on grammatical information and cost-based centering model, which are applied to various machine learning techniques. We demonstrate that our proposed features are well defined to explain pronominalization, especially pronominalization by zero pronouns in Korean, through 95 texts from three genres - Descriptive texts, News, and Short Aesop's Fables. We also show that our model significantly outperforms previous ones with a 99.9% confidence level by a T-test.