Figure 2
From: Tracing retinal vessel trees by transductive inference
Visual comparison of the segmentation result on an exemplar fundus image. (A) Original image (DRIVE-Test-Image-19), and its ground-truth segmentation in (B). (C) Segmentation result of [10]. (D) Segmentation result of [6](E) Segmentation result of [7]. (F) Segmentation result of our method. (G) Original image (DRIVE-Test-Image-01). (H) Segmentation result of [5]. (I) Segmentation result of our method. If we pay close attention to the boxed regions, we can see that our segmentation step does a much better job at retaining the connectivity of the skeleton.