Ilkwon Park

Efficient Measurement of Eye Blinking under Various Illumination Conditions for Drowsiness Detection Systems

In this paper, we propose an efficient way of measuring the level of eye blinking under various illumination conditions (such as day and night) for drowsiness detection systems which use a single camera. Determining the level of drowsiness by using eye blinking, it is an important way of detecting eye positions and measuring eyelid movements. For robust eye detection under various illumination conditions, we propose a simple illumination compensation algorithm and a novel way of measuring of eyelid movements. In order to estimate the performance of the proposed methods, we collected video data during real driving situations under various illumination conditions, such as during the day and during the night. Experimental results demonstrate an average eye detection rate of over 98% and an accurate measurement of eye blinking when using the proposed drowsiness detection system

Multi-View Video and Multi-Channel Audio Broadcasting System

In recent years, various multimedia services have become available and the demand for realistic multimedia systems is growing rapidly. The multi-view video and multi-channel audio are expected to satisfy the user demand for realistic multimedia services. In this paper, we present a new broadcasting system incorporating multi-view video and multi-channel audio over IPTV and MPEG-21 DIA. The proposed system includes data acquisition, camera calibration, data encoding and decoding, transmission, intermediate view reconstruction, and multi-view display and multi-channel audio play. In this paper, we discuss the main features of multi-view video and multi-channel audio.

Efficient Measurement of the Eye Blinking by Using Decision Function for Intelligent Vehicles

In this paper, we propose an efficient measurement of the eye blinking for drowsy driver detection system that is one of the driver safety systems for the intelligent vehicle. However, during the real driving in the daytime, drivers face is exposed to various illuminations. It makes too difficult to monitor drivers eye blinking. Therefore, we propose efficient formation of the cascaded form of Support Vector Machines (SVM) as eye verification to boost the accuracy of eye detection. Furthermore, for an efficient measurement of eye blinking, we newly define decision function that is based on the measure of eyelid movement and the weight generated by the eye classifier. In the experiments, we can show the reliable performance for our own test data acquired during a real driving in the various illumination conditions. Furthermore, through our proposed method, we use detected eye blinking for Drowsy Driver Detection System.

Interactive Multi-View Video Adaptation for 3DTV

This paper presents the descriptions related to users preferences for multi-view video presentation, which can be utilized in MPEG-21 digital item adaptation. Our description tools support the functionalities for presentation preferences and terminal capabilities. The description of presentation preferences supports users preference for the view presentation such as view selection, 3D depth adjustment, and special visual effects as well as the multi-view camera geometry conversion. Furthermore, the description of terminal capabilities supports capabilities for various display devices. Therefore, users can enjoy among monoscopic, stereoscopic, and multiscopic video they prefer. The syntax of the descriptions is represented in XML (extensible Markup Language) schema. We offer an interactive communication between DIA server and client through these descriptions related to users preferences. Their feasibility has been examined and verified in our experiments, where test sequences are adapted by view preferences.

Depth map refinement using multiple patch-based depth image completion via local stereo warping

We propose a fusion method for multiple patch-based depth image completion via local stereo warping, in order to refine the disparity map. The false disparities are often extracted from occluded areas and textureless areas in stereo matching due to their ambiguities. Our fusion method iteratively performs the following three steps in order to reconstruct the areas with false disparities. First, we generate a reliability map from the initial disparity map. Second, we extract multiple patch candidates by using the proposed depth image completion. Finally, we verify each patch candidate by using local stereo warping in order to update the reliability map. Experimental results show that the proposed method is very robust for large textureless areas and occluded areas as well because our method yields superior results in comparison to many state-of-the-art methods using stereo matching.

Efficient Data Representation of Stereo Images Using Edge-based Mesh Optimization

This paper proposes an efficient data representation of stereo images using edge-based mesh optimization. Mash-based two dimensional warping for stereo images mainly depends on the performance of a node selection and a disparity estimation of selected nodes. Therefore, the proposed method first of all constructs the feature map which consists of both strong edges and boundary lines of objects for node selection and then generates a grid-based mesh structure using initial nodes. The displacement of each nodal position is iteratively estimated by minimizing the predicted errors between target image and predicted image after two dimensional warping for local area. Generally, iterative two dimensional warping for optimized nodal position required a high time complexity. To overcome this problem, we assume that input stereo images are only horizontal disparity and that optimal nodal position is located on the edge include object boundary lines. Therefore, proposed iterative warping method performs searching process to find optimal nodal position only on edge lines along the horizontal lines. In the experiments, we compare our proposed method with the other mesh-based methods with respect to the quality by using Peak Signal to Noise Ratio (PSNR) according to the number of nodes. Furthermore, computational complexity for an optimal mesh generation is also estimated. Therefore, we have the results that our proposed method provides an efficient stereo image representation not only fast optimal mesh generation but also decreasing of quality deterioration in spite of a small number of nodes through our experiments.

Automatic background substitution using monocular camera and temporal foreground probability model

This paper describes an automatic background substitution for PC-softphone applications. Different from previous works on this topic which are mostly using stereo camera, we provide a solution which can segment the foreground region with monocular camera which is not moving. There are three main contributions from this paper. First, our proposed algorithm can segment the foreground region without any manual initialization. Second, we improve accuracy of the foreground segmentation using the Temporal Foreground Probability Model(TFPM). Finally, natural composite image was achieved by adaptive color and intensity correction method

Real-time stereoscopic conversion with adaptable viewing distance at personal stereoscopic viewing devices

This paper proposes the real-time stereoscopic conversion method with adaptable viewing distance by the viewers eye detection at personal stereoscopic viewing devices. The distance between a screen and viewer is often changed by moving a personal display or moving head position. It causes a visual discomfort and low quality stereoscopy. Our proposed method is divided into two parts; eye detection part to acquire viewers eye positions and stereoscopic generation part in real-time. In the first part, we can measure an interocular distance of a single viewer by the proposed eye detection method and also calculate the distance between a screen and a viewer by using the principle of stereoscopic generation. In the other part, we propose an efficient motion-based approach for conversion from 2D video to stereoscopic video. The disparity between left image and right image is adaptively adjusted by viewers interocular distance. Adaptation time for a 3D immersion during 3D stereoscopic viewing is remarkably reduced and stereoscopy is personally optimized. Therefore, our proposed system can offer optimized stereoscopic conversion and more freedom of viewers head considering the distance between a screen and a viewer through the proposed eye detection.

Fast gradient-based mesh generation method for the stereo image representation

This paper proposes a fast gradient-based mesh generation method for the stereo image representation. In our approach, right image in stereo image and intermediate views are fundamentally predicted and synthesized from left image by 2D image warping using regular mesh and vice versa. To overcome texture distortion on object boundaries and preserve texture smoothness of homogenous areas, we propose node selection on the strong edges in the gradient map. Furthermore, the selected nodes are evaluated by stereo matching error and validated by cross validation for nodal disparity. Each node point is iteratively moving along the pixels with high gradient value instead every point to find the optimal node position. Therefore, our approach provides a fast mesh optimization as well as a reliable image quality. The experimental results show that the proposed approach provides computational efficiency and high PSNR of prediction image compared to previous methods.

Integrating Multi-view Stereoscopic Transmission System into MPEG-21 DIA (Digital Item Adaptation)

In general multi-view system, all the view sequences acquired at the server are transmitted to the client. However, this kind of system requires high processing power of the server as well as the client, thus it is posing a difficulty in practical applications. To overcome this problem, a relatively simple method is to transmit only two view-sequences requested by the client in order to deliver a stereoscopic video. In this system, effective communication between the server and the client is one of important aspects. Therefore, we propose an efficient multi-view system that transmits two view-sequences according to users request. The view selection process is integrated into MPEG-21 DIA (Digital Item Adaptation) so that our system is compatible to MPEG-21 multimedia framework. Furthermore, multi-view descriptors related to multi-view camera and systems are newly introduced. The syntax of the descriptions and their elements is represented in XML (eXtensible Markup Language) schema. Intermediate view reconstruction (IVR) is used to reduce such discomfort with excessive disparity. Furthermore, IVR is useful for smooth transition between two stereoscopic view sequences. Finally, through the implementation of testbed, we can show the valuables and possibilities of our system.