Mao-Hsiung Hung

Event Detection of Broadcast Baseball Videos

This paper presents an effective and efficient event detection system for broadcast baseball videos. It integrates midlevel cues including scoreboard information and shot transition patterns into event classification rules. First, a simple scoreboard detection and recognition scheme is developed to extract the game status from videos. Then, a shot transition classifier is designed to obtain the shot transition patterns, which contains several novel schemes including adaptive playfield segmentation, pitch shot detection, field shot detection, as well as infield/outfield classification. The extracted midlevel cues are used to develop an event classifier based on a Bayesian Belief Network. The network is with low complexity because the number of these cues used is small, which not only improves the generalization performance of the event classifier but also reduces system complexity as well as training efforts. Using the inference results of the network, we further derive a set of classification rules to identify baseball events. The set of rules is stored in a look-up table such that the classification is only a simple table look-up operation. The proposed approach is very simple and computational efficient. More importantly, the simulation results indicate that it identifies ten significant baseball events with 95% of precision rate and 92% of recall rate, which is very promising.

Similarity retrieval of shape images based on database classification

Abstract Shape is a key visual feature used to describe image content. This paper develops a novel shape-based similarity retrieval system based on database classification which exploits the contour and interior region of a shape efficiently. In this system, the database of shape images is categorized automatically into 11 classes by a simple contour feature. In query, the contour feature of the input image is used to decide which class the query image belongs to. Then, the possible classes are selected dynamically from the database and to form candidate sets with different priority orders. Then, ART region feature is employed to compare the query with the candidate sets according to the priority order. Instead of using the original contour of a shape image directly, we employ a rough version of the original contour for the classification of shapes. The similarity test results indicate that the proposed method improves retrieval accuracy and speed significantly, as compared to ART.

Speed Up Temporal Median Filter for Background Subtraction

Temporal median filter is one of most popular background subtraction methods. However, median operation is very time-consuming which limits its applications. This paper presents a fast algorithm to reduce the computation time of the temporal median operation. By utilizing the characteristics of high correlation of adjacent frames, the fast algorithm designs a simple mechanism to check whether the median of the current frame is equal to that of the previous frame. The proposed algorithm reduces the computing frequency of median operations significantly, and the experimental results indicate it is much faster than the existing algorithms.

Generalized playfield segmentation of sport videos using color features

This paper proposes a generalized method for playfield segmentation of various sport videos. It first estimates the probability density function (pdf) of color components of an image frame. Two hill-climbing schemes, which employ two-dimensional pdf and one-dimensional pdf, respectively, are proposed for clustering. Next, a novel algorithm that utilizes the domain knowledge of sport playfields is developed to merge those clusters into four color classes: red, green, blue, and gray. Finally, a simple scheme fuses small regions into their adjacent large regions to obtain the segmentation result. The experimental results indicate that the proposed method effectively segments the playfield regions of various sport videos.

Baseball Playfield Segmentation Using Adaptive Gaussian Mixture Models

Playfield is one of main parts appearing in typical scenes of sports video. Generally, the playfield always has very distinctive attributes. For baseball, the playfield is composed of grass and soil. The colors of grass and soil are selected as a feature to segment the playfield in our work. However, playfield colors demonstrate significant variations, which may cause a large amount of segmentation errors for color-based segmentation. In this paper, we present a new method of grass-soil playfield segmentation for baseball videos based on an adaptive Gaussian mixture model. To improve segmentation accuracy, a particular GMM model is obtained by automatic training directly from sample data for each baseball game. The simulation results indicate that it can achieve very low error rates.

A new approach based on XMPP and OSGi technology to home automation on Web

As the fast development of information and Internet technologies, lots of electrical appliances started to be digitized and combined with network technologies to provide people with much more multiple services. In this study, a new approach based on XMPP and OSGi technology to home automation on Web is proposed. An application of fault-detection mechanism for smart refrigerators is demonstrated to illustrate the realization of the proposed approach. Based on the proposed approach, a communication scheme of smart appliances is developed first. Then an inference engine is developed to derive the fault of refrigerator from the status of sensors. The appliance manufactures will be benefited from the fast self-diagnosis to remotely respond. The experimental results show the feasibility and modeling techniques applicable for the management of smart appliances on web.

Rule-based Event Detection of Broadcast Baseball Videos Using Mid-level Cues

This paper presents an effective and efficient event detection system for broadcast baseball videos. It integrates mid-level cues including scoreboard information and shot transition patterns into event classification rules. First, a simple scoreboard detection and recognition scheme is developed to extract the game status from videos. Then, a shot transition classifier is designed to obtain the shot transition patterns. The extracted mid-level cues are used to develop an event classifier based on a Bayesian belief network. Using the inference results of the network, we further derive a set of classification rules to identify baseball events. The set of rules is stored in a look-up table such that the classification is only a simple table look-up operation. The simulation results indicate that it identifies ten significant baseball events with 95% of precision rate and 89% of recall rate, which is very promising.

Power Reduction of Wireless Sensor Networks Using Ant Colony Optimization

Power consumption is one of the most important problems for wireless sensor networks because of the battery limitation in each sensor. This paper presents an ant colony optimization- (ACO-) based routing algorithm to reduce power consumption. First, a grade table is built and referred to generate several possible routing paths. Then, the ACO explores these paths to reduce the power consumption of the nodes. The simulations indicate that the proposed algorithm obtains more balanced transmission among the nodes and reduces the power consumption of the routing.

Scene Classification for Baseball Videos Using Spatial and Temporal Features

Correct classification of various kinds of scenes in sport videos is essential for higher-level content analysis such as event detection. The paper presents a novel technique for the classification of the typical scenes of baseball videos. The spatial color features are employed to detect pitch scene first. The temporal features derived from a shot are utilized to classify infield and outfield scenes in the second stage. Simulation results indicate that high accuracy (more than 90 %) of classification is achieved.

An effective image enhancement method for electronic portal images

Due to the inherent low-contrast in Electronic Portal Images (EPI), the perception quality of EPI has certain gap to the expectation of most physicians. It is essential to have effective post-processing methods to enhance the visual quality of EPI. However, only limited efforts had been paid to this issue in the past decade. To this problem, an integrated approach featuring automatic thresholding is developed and presented in this article. Firstly, Gray-Level Grouping (GLG) is applied to improve the global contrast of the whole image. Secondly, Adaptive Image Contrast Enhancement (AICE) is used to refine the local contrast within a neighborhood. Finally, a simple spatial filter is employed to reduce noises. The experimental results indicate that the proposed method greatly improves the visual perceptibility as compared with previous approaches.