Koichiro Suzuki

Coded Lens: using coded aperture for low-cost and versatile imaging

We propose Coded Lens, a novel system for lensless photography. The system does not require highly calibrated optics, but instead, utilizes a coded aperture for guiding lights. Compressed sensing (CS) is used to reconstruct scene from the raw image obtained through the coded aperture. Experimenting with synthetic and real scenes, we show the applicability of the technique and also demonstrate additional functionality such as changing focus programmatically. We believe this will lead to a more compact, cheaper and even versatile imaging systems.

Sparse Isotropic Hashing

This paper address the problem of binary coding of real vectors for efficient similarity computations. It has been argued that orthogonal transformation of center-subtracted vectors followed by sign function produces binary codes which well preserve similarities in the original space, especially when orthogonally transformed vectors have covariance matrix with equal diagonal elements. We propose a simple hashing algorithm that can orthogonally transform an arbitrary covariance matrix to the one with equal diagonal elements. We further expand this method to make the projection matrix sparse, which yield faster coding. It is demonstrated that proposed methods have comparable level of similarity preservation to the existing methods.

Bias angle error self-correction for automotive applications using phased array radars installed behind bumpers

In certain automotive safety and comfort applications, phased array radars are set behind the bumpers of vehicles. Due to paint on the bumper that may contain metallic powder and to the complicated structure around installed radars, the angle detected by the radars can sometimes contain a bias error related to the signal source angle. Because the cost to calibrate radars after installation behind a bumper is high, we consider a framework of self-correction as an alternative. In this framework, there is no need to calibrate with a targets known angle ; rather, we can generate a correction table using Gaussian process MCMC simply from the radar observations.