Purchase this article with an account.
P. Liu, O. Tan, Y. Li, D. Huang; Image Quality Enhancement in Ophthalmic Optical Coherence Tomography by Frame Registration and Averaging. Invest. Ophthalmol. Vis. Sci. 2008;49(13):1838. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To evaluate quality enhancement for Fourier-domain optical coherence tomography (FD-OCT) images by averaging multiple registered frames.
7 normal subjects were scanned with a commercially available FD-OCT system (RTVue with software version 2.7, Optovue, Fremont, CA) . The system scans 26,000 a-scans per second and has 5 micron axial resolution. Subjects were scanned with line scans on retina (centered at fovea, scan length=6mm), cornea (centered at vertex, scan length=6mm) and anterior chamber angle (scan length= 4mm). Sixteen consecutive frames are acquired for each scan. A reference frame was picked by the operator and a rectangular region was drawn on the reference frame. An automatic algorithm was applied to register other frames to the content in the rectangle. The algorithm automatically determines if registration is successful for each frame. Successfully registered frames were averaged in the logarithmic domain. Contrast-to-noise ratio (CNR) enhancement was used to quantify the reduction of background and speckle noise. CNR enhancement ratio is equal to the CNR of averaged image divided by CNR of reference image. To evaluate edge preservation, a cross correlation parameter β (F. Sattar et al., "Image enhancement based on a nonlinear multiscale method," IEEE Trans. Image Processing, vol. 6, pp. 888-895, June 1997) was used. The interpretation of β is: 0 to 0.20, poor; 0.21 to 0.40, fair; 0.41 to 0.60, moderate; 0.61 to 0.80, good; and above 0.80, excellent.
The values of CNR enhancement ratio and β are shown in Table 1
The frame-averaging algorithm on the RTVue FD-OCT system successfully improved image quality by suppressing background and speckle noises. A moderate loss of edge sharpness was observed, indicating that there is room for further improvement of the registration algorithm. The algorithm worked best on corneal images.
This PDF is available to Subscribers Only