| 酒井 一樹 | 1451050: M, 2回目発表 | 加藤 博一，向川 康博，Christian Sandor，武富 貴史 |
|
title: Evaluation of accurate depth estimation using tracked feature points for dense reconstruction
abstract: In Visual SLAM method, which means tracking current camera pose and recovering the three-dimensional structure of environment in real-time, there are two ways: 1st is using exploiting image features, and 2nd is using whole pixels in frame. In the latter one, DTAM:Dense Camera Tracking and Mapping in Real-Time can estimate 3D structure for all pixels in keyframe. This method, however, produces an artifact in strong texture area. Previous work solves this problems using feature points depth to fix photometric error, but there are no quantitative evaluation. We'll present the evaluation results, and research schedule. language of the presentation: Japanese 発表題目: 密な三次元復元における特徴点制約を加えた奥行き情報の高精度な復元手法の定量評価 発表概要:リアルタイムに実環境の三次元形状とカメラの位置・姿勢情報を推定するVisual SLAMは, 画像特徴を抽出して利用する手法と,画像そのものを入力として用いる手法に大別される. 後者の中でも, キーフレーム中の全画素に対して奥行きを復元する手法として,DTAM: Dense Tracking and Mapping in Real-Timeがある.この手法は画像の同一平面上に強い輝度勾配が観測される場合は, 復元結果にアーチファクトが観測されるという問題点があった. この問題に対して, 画像中の特徴点の三次元位置情報を用いて解決する手法が提案されているが, その定量的な評価は行われていない. 本発表では,その定量評価の結果を述べ、さらに今後より精度よく復元するための研究の方向性について述べる. | ||
| 瀬戸 将志 | 1451061: M, 2回目発表 | 加藤 博一，向川 康博，Christian Sandor，武富 貴史 |
|
title: Inter-frame delay calibration free camera pose estimation on Rolling Shutter
abstract: Video acquired by handheld CMOS cameras may suffer from rolling shutter phenomenon. Rolling shutter phenomenon, which is due to the rows in the image sensor array being exposed sequentially from top to bottom, increase with the speed of the relative motion between the scene and the camera. In order to rectify rolling shutter phenomenon, previous research mainly used interpolation technique, which requires the inter -frame delay. Previous research calibrate the inter-frame delay in advance. I propose the inter-frame delay calibration free method by solving the inter-frame delay added minimizing problem. language of the presentation: Japanese | ||
| 壷内 貴弘 | 1451076: M, 2回目発表 | 加藤 博一，向川 康博，Christian Sandor，武富 貴史 |
|
title:Verification for the realization of near-eye optical-see through head worn displays based on the microlens
abstract: Recently, there has been a lot of active research and development of head worn displays. The purpose of my research examine whether MicroLens Arrays are effective in reducing the size of the Optical See Through Head Worn Displays (OST-HWD) to about the mass of regular an eye glasses. Conventional research of OST-HWDs have a problem with a focus blur of the virtual image and the augmented image. This is because the distance between the eye and the display is shorter than the eyes focal range. Therefore, I use a MicroLens Array because the virtual image and the augmented image can be displayed clearly regardless of the eyes focal limit. In this presentation, I describe how to implement a spatial image (MicroDisplay) through a MicroLens Array, and the examination of the effectiveness of OST-HWDs in the future. language of the presentation: *** Japanese *** | ||
| HUSEJIC AMAR | 1451123: M, 2回目発表 | 加藤 博一，佐藤 嘉伸，Christian Sandor，武富 貴史 |
| Title: CT Pose Deformation using One X-Ray Image for Diagnostic Assistance Abstract: In order to diagnose the problem of lower back pain doctors use X-ray and CT images. Comparing these images is hard for doctors, because X-Ray images are in 2D and offer less detail, but they can be taken in any pose (painful pose). CT image data is in 3D and offers more detailed information but can only be taken in one pose (lying down). The system that is created in our laboratory solves this problem by modifying 3D CT image data to the pose of the 2D X-Ray image. Currently the systems requires 2 X-Ray images taken in the same pose with a special X-Ray machine which all hospitals don't have. With this study we are trying to solve the problem of deformation by using only one X-Ray image. | ||