[8a-PB1-1~5] 3.3 Information photonics and image engineering

[8a-PB1-4] Microspectroscopic image analysis by filtering overcomplete basis sets after sparse coding

〇(P)Yu Nagashima^1,2, Atsushi Iwata² (1.Tokyo Univ., 2.Univ. Tokyo Hosp.)

Keywords:sparse coding, microscopic image, spectroscopy

Sparse coding has been proposed as a mathematical model for visual perception mechanisms of the spatial receptive fields in mammalian visual cortex. In the filed of image processing, it is used as a kind of independent component analysis to extract overcomplete basis sets from natural images.We implemented a new image analysis by reconstituting a set of filtered independent basis sets extracted by sparse coding of microspectroscopic images of biological samples. In this method, we could extract specific image features arising from specific biological interests without observer's intentional or unwitting bias. It is convenient that the reconstituted images preserves signal linearity of original data. This method is robust enough to handle noisy data.