Mitsuharu Matsumoto Laboratory  

Search WWW Search
Japanese | English
About us
Research projects
->Image processing
->Acoustical signal processing
->Intelligent robotics
->Interaction design
->Artificial intelligence
For students
Access map
Inside mm-labo
Image processing / Computer vision

Vision is an important sensation of humans to obtain the information from the environments.

Image processing also has an important role in digital camera and video.

Mitsuharu Matsumoto Laboratory progresses our research on image processing such as:

- Light and noise robust feature extraction from human and facial images

- Feature extraction based on edge field analysis

- Impulse noise reduction to keep high image quolity using nonlinear filters

- An optical device for close object detection

We aim to understand the mechanism of visual sensation and to apply our system to image recognition system and robot vision.


Light and noise robust feature extraction from human and facial images

Light variance and noise are important problem to be solved for face and human recognition system. It is still difficult to recognize human and faces in light variant and noisy environments although current recognition system can detect the objective human and faces from the clear images. This research aims to develop a robust feature extraction as preprocessing for recognition system and to develop a noise robust recognition system.

Feature extraction from images based on edge field analysis

We propose edge field analysis based on vector analysis and aim to extract the rotation-like feature, divergence-like feature from edge images. We especially aims to extract quasi-motion from a blurred image by using edge field analysis.

Impulse noise reduction using nonlinear filters

Although many filters for impulse noise reductions have been proposed, they reduce not only impulse noise but also signal itself. We aim to develop simple nonlinear filters, which can reduce impulse noise, while preserving edge information.

An optical device for close object detection

Although camera image is general for robot vision, it includes not only necessary information but also unnecessary information for robots. For instance, distant information in images is unnecessary when we consider the robot tasks such as object grasping and object avoidance. We, therefore, aim to develop an optical device for close object detection.

(C) Copyright 2009 - Mitsuharu Matsumoto, The university of electro-communications.