Abstract
Purpose: :
In our previous study, an automated method for reconstructing a depth map around the optic nerve head by using a stereo fundus image was investigated. In this study, a new efficient method based on the difference of the wavelength-dependent penetration depth of photographing light is proposed to obtain layer thickness by use of a stereo fundus image.
Methods: :
Color stereo fundus images taken using a white light of xenon lamp contain three color components, red (R), green (G) and blue (B). Three depth maps were generated from each component by using stereo matching technique. The depth value was calculated on the basis of the disparity between right and left image. The positions of the three depth maps are different because of different penetration depth between the R, G, and B light. Therefore, the difference of the position of the depth map was calculated as a layer thickness. Stereo retinal fundus image pairs were taken simultaneously using a commercially available stereo fundus camera (nonmydWX, Kowa Company, Ltd., Japan). The images were acquired with a digital camera attached to the stereo fundus camera. Images were 2144 x 1424 pixels in size with a 27° field of view.
Results: :
Eight stereo fundus images were taken from 8 normal subjects. The location of the depth map generated from R-, G-, and B- component in the order from shallow to deep. The layer thickness between R-G, R-B and G-B depth map were 161.1±32.7μm, 277.7±68.2μm and 116.5±46.4μm, respectively. The result could be reasonable by considering the typical thickness of the retina.
Conclusions: :
Experimental results show that the proposed technique could be useful for thickness measurement in retinal layer structure.
Keywords: image processing • nerve fiber layer • topography