From the surrounding environment, we obtain a lot of useful information. About 70% of these information are acquired through our visual system.
Apart from understanding words and pictures printed in newspapers and books, the visual system also plays a vital role in perceiving the environment to coordinate our movement (such as walking, driving and playing sports).
We are conducting researches by utilizing cameras and computers to realize human visual system as a robot's eye. |
Presently, we are performing the following topics, which include recovering of 3-D world from 2-D images, scene analysis & understanding, and visual sensing for human interface.
- Research theme
- Estimation of 3D Motion Parameters
- Estimating 3D motion (translation and rotation) parameters of object or camera from a series of images.
- 3D Shape Reconstruction from Shaded Images
- Two-image photometric stereo estimating 3D shape and surface reflectance of both smooth and polyhedral objects, and unique shape recovery of a concave polyhedron using inter reflections.
- Recognition of Handwritten Numerals by Multilayer Perceptron
- Rotation- and size-invariant recognition of handwritten numerals by using 3 layer perceptron.
- Separating Reflected Images and Real Images from Their Overlapping
- On-line separation of image reflected on the surface of glass and image of objects behind it by using optical characteristic of polarization.
- Moving Object Detection and Tracking
- A movable camera system mimicking human head and eye can detects an moving object and capture it in the center of camera view in real-time.
- Extraction of Character Strings in a Scene
- A developed system can find characters in a image by using general features of characters and character string not relying on character recognition.
- Soccer Image Sequence Computed by a Virtual Camera
- An image synthesis system which generates an image sequence from the viewpoint of a player on the ground, to create more realistic soccer scene from TV images.