Purchase this article with an account.
Luis A V Carvalho, Danilo Motta, Valeria Carvalho; Development of a Focus Stacking Algorithm for Improvement of 3D Retinal Image Reconstruction. Invest. Ophthalmol. Vis. Sci. 2015;56(7 ):5256.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
There is an urgent need in ophthalmology for more cost effective instrumentation for the early diagnosis of Glaucoma, a leading disease in the cause of blindness worldwide. In a previous project we have developed a low cost 3D retinal camera , which presented promising results. Nevertheless, one of the great challenges in 3D reconstruction is the precision with which the cross correlation of stereo images is implemented, where in turn the depth of field (DOF) also plays an important role.
In this work we have applied a focus stacking technique to images acquired with a 3D retinal camera developed in a previous project . Stereo videos were acquired from a model eye for progressive focusing planes of the object conjugated with a CCD plane with zoom lens. A matlab algorithm was developed in order to pre-process the video files and select specific image couples that had very distinct focus characteristics. In-focus regions of each image were detected automatically via edge detection and Fourier analysis implemented in our in-house matlab algorithm. The in-focus patches were then blended together to generate the final image.
Our stacking algorithm was tested on a model eye for 3 distinct optic disc shapes manufactured with precision of 5 microns. Results for the same algorithm developed in the previous phase  were compared both for non focus stacking images and also after the stacking procedure. The RMSE for all model eyes without stacking was 0.14 mm and with stacking was 0.10 mm, almost 30% more precise.
From the results obtained for this preliminary study on model eyes the conclusion is that the stacking technique, originally used in photography and microscopy, can be applied to 3D optic disc reconstruction as one more factor for precision improvement. Our goal in the next phase is to apply the same algorithms developed here to in vivo eyes and compare results with other more sophisticated (and more expensive) optic nerve depth analysis techniques, such as SLO.<br /> <br />  CARVALHO LAV, RAMOS E, ROMANO A, Construction of a True-depth Quantitative and Low cost 3d Optical Disc Measurement Device and Collaborative Diagnosis using Stereo Images and Videos. In: ARVO, Invest Ophthalmol Vis Sci 2012; 53: E-Abstract 3102, 2012.<br />
This PDF is available to Subscribers Only