Research Projects

Snapshot multispectral polarization imaging

A new filter array and demosaicking method for snapshot multispectral polarization imaging is proposed. The multispectral polarization filter array is developed by forming the photonic crystal structure at each pixel. We propose a demosaicking method for multispectral polarization images based on snapshot imaging using a linear model. The results show that the proposed imaging system, which is a simple demosaicking device, can recover a wide region of the spectrum, including visible wavelengths, in a single shot.

Multispectral filter array and demosaicking

Multispectral images have been studied for applications in fields such as remote sensing, industry, agriculture, medicine, and consumer products. We propose a design method for multispectral filter arrays (MSFAs) based on an RGB color filter array. We aim to develop a high-speed, low-cost, precise multispectral imaging system by utilizing both an MSFA pattern and demosaicking.

Computer-aided diagnosis in digital pathology

Pathologists specialize in interpreting laboratory tests and evaluating cells, tissues, and organs to diagnose disease. Digitizing glass pathology slides can make the diagnostic process much faster and easier for pathologists. We propose methods for multispectral imaging, digital staining, and quantification of pathological tissues for computer-aided diagnosis.

Quality measurements of fruits and vegetables

The internal properties of fruits and vegetables, such as maturity, sugar content, and acidity, can be used to quantify their taste. The use of near-infrared imaging as a rapid nondestructive technique for measuring internal properties has the potential to significantly increase productivity, safety, and quality consistency. We propose a method that captures hyperspectral images of food and calculates the potential map of their sugar and moisture content.

Embedding spectral information into RGB images

Most digital imaging techniques utilize the RGB color space, making it difficult to realize multispectral imaging applications. We propose a method that embeds multispectral data into an RGB image to facilitate sharing of multispectral images. The embedded multispectral image can be reconstructed from the generated RGB image, which can also be used as a normal RGB image by extracting the multispectral data.

Next-generation image compression

The widespread use of digital images has led to great demand for more efficient image compression. We propose compression algorithms that target the region of interest in an image or raw image data before demosaicking.