Che-Hao Hsu

Computer-aided classification of lung nodules on computed tomography images via deep learning technique

Lung cancer has a poor prognosis when not diagnosed early and unresectable lesions are present. The management of small lung nodules noted on computed tomography scan is controversial due to uncertain tumor characteristics. A conventional computer-aided diagnosis (CAD) scheme requires several image processing and pattern recognition steps to accomplish a quantitative tumor differentiation result. In such an ad hoc image analysis pipeline, every step depends heavily on the performance of the previous step. Accordingly, tuning of classification performance in a conventional CAD scheme is very complicated and arduous. Deep learning techniques, on the other hand, have the intrinsic advantage of an automatic exploitation feature and tuning of performance in a seamless fashion. In this study, we attempted to simplify the image analysis pipeline of conventional CAD with deep learning techniques. Specifically, we introduced models of a deep belief network and a convolutional neural network in the context of nodule classification in computed tomography images. Two baseline methods with feature computing steps were implemented for comparison. The experimental results suggest that deep learning methods could achieve better discriminative results and hold promise in the CAD application domain.

HoloTube: a low-cost portable 360-degree interactive autostereoscopic display

This paper proposes a novel, low cost, and portable 360-degree cylindrical interactive autostereoscopic 3D display system. The proposed system consists of three parts: the optical architecture (for back-projecting image correctly on the cylindrical screen), the projection image transformation workflow (for image rectifying and generating multi-view images), and the 360-degree motion detection module (for identifying viewers’ locations and providing the corresponding views). Based on the proposed design, only one commercial micro projector is employed for the proposed cylindrical screen. The proposed display offers great depth perception (stereoacuity) with a special designed thick barrier sheet attached to the screen. The viewers are not required to wear special glasses and within appropriate range (< 5m) the viewers can view the screen at any distance and angle. The user study verified that the proposed display offers satisfactory depth perception (binocular parallax, shading distribution, and linear perspective) for various viewing distances and angles without noticeable discomfort. The production cost of the current prototype is about USD

CrossbowCam: a handheld adjustable multi-camera system

300. With mass production, the unit cost is expected to decline to within USD

HoloTabletop: an anamorphic illusion interactive holographic-like tabletop system

60. The proposed display system has the advantages of ease of use, low production cost, high portability and mobility. The proposed system is suitable for application such as museum virtual exhibition, remote meeting, multi-user online game, etc. We believe that the proposed system is very promising for the market of low-cost portable 360-degree interactive autosereoscopic displays.

Omni-Tube: a low-cost portable omnidirectional interactive 3D display

This paper presents a novel multi-functional, low-cost handheld multi-camera system (one dimensional camera array) - “CrossbowCam”. The CrossbowCam is suitable for multi-viewpoint image acquisition, smooth switching, alignment and seamless stitching applications. The proposed system differs from the traditional fixed image acquisition systems which are large-sized, high-priced, single functional, and can only captured images at specific locations. With the proposed system, the users can push one single button to change the configuration of the camera array rapidly to divergence (convex arc), parallel (linear), or convergence (concave arc). The three camera configurations can each be suitable for applications such as panorama image stitching, autostereoscopic 3D display, bullet-time (time-freeze) visual effect, 3D scene reconstruction, etc. To rapidly acquire the relationship among cameras after configuration change, we propose a two-stage calibration method to compensate the mechanical misalignment. The first stage adopts the traditional checkerboard calibration method to get the intrinsic parameters (focal length, principal point) and the lens distortion for each camera. The second stage requires no auxiliary tool but utilizes a large number of common feature points from multiple viewpoint images to acquire the extrinsic parameters (translation and rotation matrix) and to compensate the vertical misalignment and the horizontal uneven angle distribution due to the mechanical structure. The proposed system can then insert virtual viewpoint images between actual viewpoint images to allow the viewpoint switching more smoothly. The proposed system has eight cameras with maximum viewing angle of 90° in divergence mode, 38 mm spacing in parallel mode, and imaging radius of 10 m ∼ 0.5 m in convergence mode. We believe that the proposed system can potentially change the consumer habits and becomes the new type of home-use handheld camcorder system in the future.

Multicast scheduling for stereoscopic video in wireless networks

HoloTabletop is a low-cost holographic-like tabletop interactive system. This system analyzes user’s head position and gaze location in a real time setting and computes the corresponding anamorphic illusion image. The anamorphic illusion image is displayed on a 2D horizontally-located monitor, yet offers stereo vision to the user. The user is able to view and interact with the 3D virtual objects without wearing any special glasses or devices. The experimental results and user studies verify that the proposed HoloTabletop system offers excellent stereo vision while no visual fatigue will be caused to human eyes. This system is a great solution for many interactive applications such as 3D board games and stereo map browsing.

An efficient algorithm for periodic halftone identification

Flat panel autostereoscopic 3D display is a mature and production-ready technology which employs specially designed optical devices, such as lenticular lens arrays, parallax barrier, and directional backlight. To overcome the restricted range of viewing angles and distances of these approaches, SONY RayModeler [Ito et al. 2010] and Osaka university [Yagi et al. 2011] have developed new omnidirectional displays. However, their approaches require the use of high cost equipments, such as high-speed rotary motor with synchronous control system, 2D LED dot matrix, high-speed multi-projector, fog screen generator, etc. In this work, we showcase the design and implementation of a novel 3D display, Omni-Tube. It is designed to be low-cost, compact and easy to use. It can be easily mounted to any commercial micro projector, hence it is suitable for various applications, such as virtual interactive exhibition in a museum or during a conference meeting.

A mixed-reality showcase for multiple users from unconstrained viewing angles

Stereoscopic video multicast over wireless network is a challenging issue due to large bandwidth requirement, limited resource, and heterogeneous user channel conditions. Recently, most existing methods for stereoscopic video multicast employ symmetric video coding that transmits the same video quality for stereo views. In this paper, we propose a novel rate scheduling method for stereoscopic video multicast in WiFi networks through asymmetric video coding to maximize users’ perceived video quality. We first formulated rate scheduling problem which has complexity in non-polynomial time subjected to playback time limit, block dependency, and the ratio of asymmetric video quality for stereo views. Then, a novel algorithm is proposed to assign a suitable rate for each frame per layer. Furthermore, we studied the impact of block dependency and asymmetric coding. Experimental results confirm that our approach resulted in promising perceived video quality while outperforming several existing video multicast techniques.

USER INTERFACE OF AN ELECTRONIC DEVICE

When periodic halftoned documents (e.g., books, newspapers, and magazines) are scanned for image reproduction, moiré patterns occur. In order to avoid these moiré artifacts, it is necessary to detect the periodic halftone. This paper provides a fast Fourier transform based method to classify periodic halftoned documents. Experimental results show that the overall accuracy of this method is 97% on a large data set which contains many difficult-to-classify images. Misclassified documents tend to be extremely difficult to classify, in that they contain very small periodic halftone regions. An existing method, by comparison, has accuracy of only 70%.

Physiognomy master: a novel personality analysis system based on facial features

An ideal showcase should allow its viewers not only see the display object clearly, but also obtain sufficient information of the object by static texts or dynamic images. This extra information will enhance the impression the display object makes on viewers. For a 360-degree showcase, it is a challenging problem to provide proper virtual images for users at different viewing locations. In literature, [Kim et al.] proposed a pyramid-shaped showcase in which the display object is placed and half-silvered mirrors were employed to reflect the virtual image for four viewing directions (left, right, front, and rear). Though the user can observe both the virtual image and real object well in each of the aforementioned viewing directions, the geometry inconsistency would occur at the middle of the consecutive two viewing directions. Commercial product, Lite-Fast MAGIC [Kinoton], also developed a 360-degree mixed-reality transparent LED display, however it will result in ghost artifact when showing virtual information for different viewing directions due to its transparent material.