Translator Disclaimer
19 April 2017 Optimization method of superpixel analysis for multi-contrast Jones matrix tomography (Conference Presentation)
Author Affiliations +
Abstract
Local statistics are widely utilized for quantification and image processing of OCT. For example, local mean is used to reduce speckle, local variation of polarization state (degree-of-polarization-uniformity (DOPU)) is used to visualize melanin. Conventionally, these statistics are calculated in a rectangle kernel whose size is uniform over the image. However, the fixed size and shape of the kernel result in a tradeoff between image sharpness and statistical accuracy. Superpixel is a cluster of pixels which is generated by grouping image pixels based on the spatial proximity and similarity of signal values. Superpixels have variant size and flexible shapes which preserve the tissue structure. Here we demonstrate a new superpixel method which is tailored for multifunctional Jones matrix OCT (JM-OCT). This new method forms the superpixels by clustering image pixels in a 6-dimensional (6-D) feature space (spatial two dimensions and four dimensions of optical features). All image pixels were clustered based on their spatial proximity and optical feature similarity. The optical features are scattering, OCT-A, birefringence and DOPU. The method is applied to retinal OCT. Generated superpixels preserve the tissue structures such as retinal layers, sclera, vessels, and retinal pigment epithelium. Hence, superpixel can be utilized as a local statistics kernel which would be more suitable than a uniform rectangle kernel. Superpixelized image also can be used for further image processing and analysis. Since it reduces the number of pixels to be analyzed, it reduce the computational cost of such image processing.
Conference Presentation
© (2017) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Arata Miyazawa, Young-Joo Hong, Shuichi Makita, Deepa K. Kasaragod, Masahiro Miura, and Yoshiaki Yasuno "Optimization method of superpixel analysis for multi-contrast Jones matrix tomography (Conference Presentation)", Proc. SPIE 10053, Optical Coherence Tomography and Coherence Domain Optical Methods in Biomedicine XXI, 1005307 (19 April 2017); doi: 10.1117/12.2254281; https://doi.org/10.1117/12.2254281
PROCEEDINGS
PRESENTATION ONLY


SHARE
Advertisement
Advertisement
Back to Top