Il Kwon Jeong

Fast focus estimation using frequency analysis in digital holography

A novel fast frequency-based method to estimate the focus distance of digital hologram for a single object is proposed. The focus distance is computed by analyzing the distribution of intersections of smoothed-rays. The smoothed-rays are determined by the directions of energy flow which are computed from local spatial frequency spectrum based on the windowed Fourier transform. So our method uses only the intrinsic frequency information of the optical field on the hologram and therefore does not require any sequential numerical reconstructions and focus detection techniques of conventional photography, both of which are the essential parts in previous methods. To show the effectiveness of our method, numerical results and analysis are presented as well.

Cylindrical angular spectrum using Fourier coefficients of point light source and its application to fast hologram calculation

We will introduce a new simple analytic formula of the Fourier coefficient of the 3D field distribution of a point light source to generate a cylindrical angular spectrum which captures the object wave in 360° in the 3D Fourier space. Conceptually, the cylindrical angular spectrum can be understood as a cylindrical version of the omnidirectional spectral approach of Sando et al. Our Fourier coefficient formula is based on an intuitive observation that a point light radiates uniformly in all directions. Our formula is defined over all frequency vectors lying on the entire sphere in the 3D Fourier space and is more natural and computationally more efficient for all around recording of the object wave than that of the previous omnidirectional spectral method. A generalized frequency-based occlusion culling method for an arbitrary complex object is also proposed to enhance the 3D quality of a hologram. As a practical application of the cylindrical angular spectrum, an interactive hologram example is presented together with implementation details.

Increasing the depth of field for multiview 3D images

One of the overarching goals for 3D displays is allowing viewers to perceive depth. This can be achieved through ‘continuous parallax,’ which is provided by the displays themselves. Indeed, the concept of continuous parallax is the principal factor that governs the creation of natural viewing conditions with a 3D display. It is also the reason why eyes do not suffer from the vergence-accommodation conflict when people view natural scenes and objects. Continuous parallax can be achieved with a type of 3D display known as a ‘super-multiview display’ (a concept first introduced in the 1990s1). In super-multiview displays it is necessary for at least two different image views to be provided simultaneously to each eye of a viewer. Through this process, it is possible for a viewer to perceive a hologram with a sense of depth. In other words, a monocular sense of depth is achieved by simultaneously projecting two different-view images to each pupil of a viewer’s eyes. Although the realization of a monocular sense of depth, with 26 continuously projected different-view images directed to the eyes of viewers, was reported in 2011,2 no additional information on this subject has since been published. We have therefore been investigating ways to verify the effects of increasing the number of simultaneously projected differentview images to each eye of a subject. As part of this work we have developed a super-multiview condition simulator that can be used to project up to four different-view images to each eye simultaneously.3 With our simulator we can address a number of important questions that need to be answered before a commercial super-multiview display is built. For instance, is it possible to obtain a monocular sense of depth with more than two simultaneously projected images? In addition, does the focusable depth range increase proportionally with the number of different-view images that are simultaneously projected to the pupils? Figure 1. Schematic illustration of the super-multiview effect simulator concept. Up to four different-view images can be simultaneously projected onto each pupil of a viewer. The depth of field (DOF) of the different projection configurations, expressed in units of diopter (D), is known for stereo image and two different-view image situations, but has not previously been obtained for three or four different-view projections.

Volume matting: object tracking based matting tool

Digital matting, whose goal is to extract only an interesting foreground component from arbitrary and natural background regions, plays an important role in a variety of areas such as computer vision, graphics and so on. Especially, digital matting has been widely used in film production lately for providing an efficient way of dealing with a complicated composition. However, it is hard to generate a perfect matte from a given image without any prior information because the matting problem is intrinsically ill-posed. This prior information is usually fed by means of a trimap or scribbles. Hence, when extracting foreground objects from any sequent images with the still-image matting method, the works would be definitely laborious and time-consuming. To overcome these problems, video matting of the process of pulling a foreground regions from sequent images has been introduced. For instance, the video matting method proposed by [Chuang et al. 2002] leverages a bayesian technique for image matting and an optical flow algorithm to estimate the trimap flow.

Interactive Computer-generated Hologram Using Spherical Angular Spectrum

We will present an interactive computer-generated hologram technique using spherical angular spectrum. The spherical angular spectrum consists of the frequency spectrum of the object wave in all-around directions. The spherical angular spectrum is computed by a formula of the Fourier coefficients of point light source. Given a spherical angular spectrum, one can generate a planar hologram at a high speed for any given direction. We also present an interactive computer-generated hologram example using a hand-recognition sensor.

LiveTree: realistic tree growth simulation tool

Natural 3D tree modeling and growth simulation as realistic as possible have been an important goal in a variety of areas such as computer vision and graphics. Hence, the creation of realistic looking 3D tree growth models is one of the most complicated and challenging tasks because of its inherent geometric complexity. As a consequence, computer representations of tree growth processes need considerable efforts to achieve high level of realism. Moreover, when describing the growth processes for a given whole tree model in most of commercial tools, only the scaling factor of overall tree size is considered so that the tree growth simulation has a limitation of having a simple isometric growth, that is the trees shape is consistent through the entire growth processes. In reality, meanwhile, every parts of a tree grow at relative rates.