Beom-Joon Cho

Locating characters in scene images using frequency features

This paper presents a (language-independent) method of locating rectangular text regions in natural scene images. The method consists of two steps that can be applied in succession or independently: the frequency of edge pixels across vertical and horizontal scan lines, and the fundamental frequency in the Fourier domain. The frequency feature of text images is highly intuitive, and this is the focus of the research. The detection of rectangles using a Hough transform is also addressed. Texts that are meaningful to many viewers usually appear in rectangles of colours of high contrast to the background. Hence it is natural to assume that the detection of rectangles may be helpful for locating desired texts correctly in natural outdoor scene images.

A layered scripting language technique for avatar behavior representation and control

The paper proposes a layered scripting language technique for representation and control of avatar behavior for simpler avatar control in various domain environments. We suggest three layered architecture which is consisted of task-level behavior, high-level motion, and primitive motion script language. These layers brides gap between application domain and implementation environments, so that end user can control the avatar through easy and simple task-level scripting language without concerning low-level animation and the script can be applied various implementations regardless of application domain types. Our goal is to support flexible and extensible representation and control of avatar behavior by layered approach separating application domains and implementation tools.

Two Texture Segmentation of Document Image Using Wavelet Packet Analysis

In this paper, we present a text segmentation method using wavelet packet analysis and k-means clustering algorithm. This approach assumes that the text and non-text regions are considered as two different texture regions. The text segmentation is achieved by using wavelet packet analysis as a feature analysis. The wavelet packet analysis is a method of wavelet decomposition that offers a richer range of possibilities for document image. From these multiscale features, we compute the local energy and intensify the features before adapting the k-means clustering algorithm based on the unsupervised learning rule. The results show that our text segmentation method is effective for document images scanned from newspapers and journals.

Facial component features for facial expression identification using Active Appearance Models

Statistical methods of shape and appearance are powerful tools used in computer vision for near-correct interpretation of images. In this paper, we present a method for classifying facial expressions based on the extracted features of facial components. The face, the window to the inner self of an individual can be analyzed for outright expressions like sadness, happiness, anger, surprise, disgust and fear. The facial region is detected; pre-processing is done on the image by using Active Appearance Model (AAM) to extract the vital feature on the facial components. Six classes are formed based on the different expressions and the model from the AAM procedure used to compare with the query image/face. Mahalanobis distance algorithm is used to classify the image in question into the best-fit class. Japanese Female Facial Expression (JAFFE) public database is used to evaluate our method with over 200 images of still images used in our experiment. A higher classification rate observed.

Efficient Segmentation Path Generation for Unconstrained Handwritten Hangul Character

This study suggests background thinning method for segmenting character unit of handwritten Hangul. Background thinning method conducts thinning processing using background information between characters and shows effective performance in segmenting for overlapped and touched characters. Character segmentation method using background thinning shows rapid segmentation performance with external segmentation which needs no judgment of recognition process. This method showed excellent performance in touched character segmentation as well as in segmentation of overlapped characters.

Silhouette-Edge-Based Descriptor for Human Action Representation and Recognition

Extraction and representation of postures and/or gestures from human activities in videos have been a focus of research in this area of action recognition. With various applications cropping up from different fields, this paper seeks to improve the performance of these action recognition machines by proposing a shape-based silhouette-edge descriptor for the human body. Information entropy, a method to measure the randomness of a sequence of symbols, is used to aid the selection of vital key postures from video frames. Morphological operations are applied to extract and stack edges to uniquely represent different actions shape-wise. To classify an action from a new input video, a Hausdorff distance measure is applied between the gallery representations and the query images formed from the proposed procedure. The method is tested on known public databases for its validation. An effective method of human action annotation and description has been effectively achieved.

Windows Security Patch Auto-Management System Based on XML

In recent days, damages to information systems and network due to worm and virus using vulnerabilities of windows security have been rapidly increasing. How to deal with the attack using vulnerabilities of windows program is to install patch appropriately and rapidly. This study suggests security patch auto-management system which installs security patch file automatically to clients through automatic downloading of the patch from MS download center based on XML as existing patch management system needs intervention of managers.

Genetic algorithm-based video segmentation with adaptive population size

This paper presented a novel object-based video segmentation using genetic algorithms. The novelty of the approach is that the population size is not constant, but motion dependent. The population size depends on the degree of the motion. In our approach, the video segmentation is performed by two steps: initial segmentation and temporal tracking. Once the objects constituting the scenes, which are tracked through the whole video sequence. Then, the temporal tracking is carried out by chromosomes that evolve using distributed genetic algorithms (DGAs). Each chromosome has its own population size according to its motion amounts, and independently evolves using local evolutionary rules. The proposed method was tested with well-known video sequences, and the results confirmed its effectiveness in segmenting a video sequence.

Global Feature Extraction and Recognition from Matrices of Gabor Feature Faces

This paper presents a method for facial feature representation and recognition from the Covariance Matrices of the Gabor-filtered images. Gabor filters are a very powerful tool for processing images that respond to different local orientations and wave numbers around points of interest, especially on the local features on the face. This is a very unique attribute needed to extract special features around the facial components like eyebrows, eyes, mouth and nose. The Covariance matrices computed on Gabor filtered faces are adopted as the feature representation for face recognition. Geodesic distance measure is used as a matching measure and is preferred for its global consistency over other methods. Geodesic measure takes into consideration the position of the data points in addition to the geometric structure of given face images. The proposed method is invariant and robust under rotation, pose, or boundary distortion. Tests run on random images and also on publicly available JAFFE and FRAV3D face recognition databases provide impressively high percentage of recognition.

Facial Expression Classification through Covariance Matrix Correlations

This paper attempts to classify known facial expressions and to establish the correlations between two regions (eye + eyebrows and mouth) in identifying the six prototypic expressions. Covariance is used to describe region texture that captures facial features for classification. The texture captured exhibit the pattern observed during the execution of particular expressions. Feature matching is done by simple distance measure between the probe and the modeled representations of eye and mouth components. We target JAFFE database in this experiment to validate our claim. A high classification rate is observed from the mouth component and the correlation between the two (eye and mouth) components. Eye component exhibits a lower classification rate if used independently.