Faculty of Engineering and Natural Sciences

Advanced Topics in Communications, Multimedia

and Networking

A Seminar Series 
 
by
 
Ass. Prof. Yücel Altunbaşak
 School of Electrical and Computer Engineering
Georgia Institute of Technology
 

Seminar 3:

Color plane interpolation using alternating projections

&

Hyperspectral Image Modeling and Reconstruction

May 28, 2003, 16:10, FENS G035

Part-I: “Color plane interpolation using alternating projections”

Most commercia1 digital cameras use color filter arrays to sample red, green, and blue colors according to a specific pattern. At the location of each pixel only one color sample is taken, and the values of the other colors must be interpolated using neighboring samples. This color plane interpolation is known as demosaicing; it is one of the important tasks in a digital camera pipeline. If demosaicing is not performed appropriately, images suffer from highly visible color artifacts. In this talk we present a new demosaicing technique that uses inter-channel correlation effectively in an alternating-projections scheme. We define two types of constraint sets; one imposes consistency with the observed data, and the other arises from the similarity between the high-frequency components of the color channels. An initial estimate is projected onto these constraint sets iteratively until convergence is achieved. We have compared this technique with six state-of-the-art demosaicing techniques, and it outperforms all of them, both visually and in terms of mean square error.

Part-II: “Hyperspectral Image Modeling and Reconstruction”

Hyperspectral images are the data type obtained for space imagery applications like target detection, tracking, agriculture, mine and oil exploration. Unfortunately, there are various effects (atmospheric scattering, secondary illumination, changing viewing angle, sensor noise just to name a few) that degrade the acquired image quality. Since resolution is one of the key parameters in a space imagery application, its improvement pays off greatly. Application of super resolution techniques separately to every spectral band is problematic, because of two reasons. First, the number of spectral bands can go up to hundreds, thus increasing the computational load excessively. Second, considering bands separately does not make use of the information that is present in all bands. In this talk, I will introduce a novel super resolution method for hyperspectral images. An integral part of our work is to model the hyperspectral image acquisition process. We propose a model that enables us to represent the hyperspectral observations from different wavelengths as weighted linear combinations of a small number of basis image planes. Then a method for applying super resolution to hyperspectral images by using this model is presented. The method fuses information from multiple frames and spectral bands to improve spatial resolution and reconstruct the spectrum of the observed scene as a combination of a small number of spectral basis functions.