Abstract

I have the expertise and knowledge to handle everything from mathematical modeling with applied mathematics to numerical analysis, software, and hardware implementation. In particular, I have a unique advantage in numerically stable implementation (computational science / numerical analysis) for estimating the dense motion distribution (optical flow / scene flow) of all subjects in a video based on the variational method.
Since there is no non-linear operation and there are no variables just for numerical calculation, it is easy to make hardware.
The research theme is a method of stably estimating the three-dimensional motion (scene flow) of the subject surface from stereo images of in-vehicle cameras and range sensor images.
For 4D images such as medical images, it is possible to estimate dense 3D motion for each 3D data point. By using the estimated motion, it can be applied to predict the course of the subject, classify by motion, improve the data compression rate of moving images, and so on.
I am familiar with the latest C++ language standards and the internal implementation of the OpenCV library, and have experience in projects related to high-speed image processing when I belonged to a company, so I can provide high-quality technical guidance.

Specific examples

Short-term future prediction, video coding

Future prospects

Estimating with special sensors, speeding up, high accuracy, etc.

Research facilities and equipments

Event camera DAVIS346, intel realsense d435i, high performance computer, GeForce RTX3090 etc.

Collaboration with external organizations

Medical image application, autonomous driving, video coding, etc.

Related patents/papers

See research publications