Early-career research(J-PEAKS)

With the widespread use of mobile phones, digital cameras have become very familiar. Various information about objects can be obtained from images taken by digital cameras. In conventional image processing fields, it is possible to estimate the color and three-dimensional shape of objects by acquiring and analyzing light directly reflected from the surface of the subject (direct light) when illuminated. However, this is limited to estimating only the surface characteristics of the object.

On the other hand, we have focused on the fact that the deep information of scenes that cannot be directly seen by the camera, such as parts located inside the object or parts hidden by occlusion, is not contained in the direct light reflected from the surface, but rather in the light that has travelled through complex paths (indirect light). Therefore, this research aims to visualize potential image information of scenes that have not been directly visible so far by selectively observing the indirect light component among the light rays propagating through the scene in a complex way.

Furthermore, focusing on the fact that light is a multidimensional phenomenon with various physical properties such as wavelength and polarization, we selectively observe and analyze the transitions of wavelength and polarization as indirect light propagates through the scene. Through this, we aim to develop a multidimensional indirect vision focusing on indirect light, which differs from conventional scene understanding and analysis using direct light.

As applications of this research, we envision developments in medical image processing by visualizing changes in the internal state of the skin and the development of in-vehicle vision systems for detecting invisible pedestrians by visualizing images beyond occluding objects. These technologies have the potential to bring innovation to various fields of society, such as medicine and traffic safety.

Click to enlarge

Hiroyuki KUBO's Article / News Release