Do-kyoon Kim

Stereoscopic conversion of monoscopic video by the transformation of vertical-to-horizontal disparity

We present a new method for converting monoscopic video to stereoscopic video. The key characteristic of our proposed method is that it can process non-horizontal camera/object motion existing in most of image scenes. It is well known that the non-horizontal motion causes the vertical parallax to human eyes and accordingly visual discomfort. The proposed methodology is composed of four major steps. First, given a current video frame, we estimate a motion vector for each block by a conventional block matching motion estimation algorithm. The motion vector is composed of horizontal and vertical disparities. Second, the norm of the motion vector is computed for each block. Here, the vertical disparity is eliminated due to the usage of the norm of the motion vector. Due to the unreliability of estimated motion vectors, a low-pass filter is performed on the norm of the motion vector in order to enhance the reliability. Third, each block is shifted to the horizontal direction by the norm of the motion vector, which is transformed to binocular parallax. The shift of blocks in the horizontal direction eliminates the effects of the vertical disparity. Finally, all the shifted blocks are synthesized and a synthesized image is then generated. A stereoscopic image pair being composed of the original image and its associated synthesized image is produced. With proper 3D viewing devices, users can feel 3D depth from seeing the stereoscopic image. Preliminary experiments have demonstrated that stable stereoscopic image pairs can be produced by applying our proposed method to a variety of monoscopic video with non-horizontal camera panning and/or object motion.

IrCube tracker: an optical 6-DOF tracker based on LED directivity

Six-degrees-of-freedom (6-DOF) trackers, which were mainly for professional computer applications, are now in demand by everyday consumer applications. With the requirements of consumer electronics in mind, we designed an optical 6-DOF tracker where a few photo-sensors can track the position and orientation of an LED cluster. The operating principle of the tracker is basically source localization by solving an inverse problem. We implemented a prototype system for a TV viewing environment, verified the feasibility of the operating principle, and evaluated the basic performance of the prototype system in terms of accuracy and speed. We also examined its application possibility to different environments, such as a tabletop computer, a tablet computer, and a mobile spatial interaction environment.

Controlled growth of CdSe quantum dots on silica spheres

Various sizes of CdSe quantum dots have been fabricated on the surface of the monodisperse silica spheres and five diffe rent photoluminescence (PL) peaks are observed from the CdSe quantum dots. The monodisperse silica spheres were syn thesized with Stöber synthetic method. The surface of the spheres was modified with 100:1 ratio of phenylpropyltrimeth oxysilane (PTMS) and mercaptopropyltrimethoxysilane (MPTMS). The MPTMS works as a covalent bond formation wi th CdSe quantum dots, and the PTMS acts as a separating quantum dots to prevent PL quenching by neighboring quantu m dots. The Fourier transform infrared (FTIR) spectrum of the surface modified spheres (SMSiO2) shows strong absorpti on peak at 2852 and 2953 cm-1 representing the characteristic absorption of –CH or -CH2. The FTIR absorption peak at 1 741 cm-1 represents the characteristic absorption of CdSe quantum dots. The field emission scanning electron microscope image shows the average diameter of the spheres ranging approximately 418 nm. The ultraviolet-visible transmittance s pectrum shows stop band at 880 nm. The PL spectrum shows five different emission bands at 434, 451, 468, 492 and 545 nm, which indicates the formation of several different sizes of CdSe quantum dots.

Motion stereo based on Fourier local phase adaptive matching

A computationally efficient pyramid adaptive algorithm based on Fourier local phase matching is proposed to compute precise depth estimates for motion stereo. In general, the motion stereo uses intensity value as the matching primitive. To decrease the computational depth error, when the brightness of each moving picture is highly variable, we developed a method for motion stereo using the phase of Gabor filters instead of intensity. To implement the motion stereo more efficiently, a hierarchical method is also used, consisting of pyramidal data se- quences. We present the experimental results demonstrating that our method gives more accurate depth estimates than the motion stereo algorithm based on intensity matching.

Combining 3D structure of real video and synthetic objects

This paper presents a new approach of combining real video and synthetic objects. The purpose of this work is to use the proposed technology in the fields of advanced animation, virtual reality, games, and so forth. Computer graphics has been used in the fields previously mentioned. Recently, some applications have added real video to graphic scenes for the purpose of augmenting the realism that the computer graphics lacks in. This approach called augmented or mixed reality can produce more realistic environment that the entire use of computer graphics. Our approach differs from the virtual reality and augmented reality in the manner that computer- generated graphic objects are combined to 3D structure extracted from monocular image sequences. The extraction of the 3D structure requires the estimation of 3D depth followed by the construction of a height map. Graphic objects are then combined to the height map. The realization of our proposed approach is carried out in the following steps: (1) We derive 3D structure from test image sequences. The extraction of the 3D structure requires the estimation of depth and the construction of a height map. Due to the contents of the test sequence, the height map represents the 3D structure. (2) The height map is modeled by Delaunay triangulation or Bezier surface and each planar surface is texture-mapped. (3) Finally, graphic objects are combined to the height map. Because 3D structure of the height map is already known, Step (3) is easily manipulated. Following this procedure, we produced an animation video demonstrating the combination of the 3D structure and graphic models. Users can navigate the realistic 3D world whose associated image is rendered on the display monitor.

Image compression based on motion segmentation

Humans subjectively evaluate the content of a scene. FOr content-based indexing and retrieval, we index and retrieve the scene containing moving objects because they remain in our memory longer than static scene. The importance of processing moving objects has been demonstrated in image compression, content-based data processing, and a variety of video processing techniques. This paper proposes the method of segmenting and then compressing the image including moving objects. The image scene is usually composed of the motion region (MR) and static region (SR). For simplicity, the camera motion region is assigned to the SR, because the region has similar characteristics with the SR. The MR is extracted by our segmentation technique. We propose a line scan-based segmentation method being composed of motion estimation and label assignment. The dominant region owning the largest number of blocks with the same label is classified as SR. The region excluding the SR is MR. Then, the MR is processed by lossless compression or lossy with low-compression ratio to preserve the high quality, and the SR by a lossy method with high compression ratio. Rather than applying separate methods to MR and SR, we use a hybrid compression method based on DCT. Experiments on test video clips show the increase of the compression ratio with respect to the lossless compression and better visualization of the moving objects compared with the lossy compression.