Robust Detection Model of Vascular Landmarks for Retinal Image Registration: A Two-Stage Convolutional Neural Network

Figure 5

The registration results for the retinal images of the S (a–f), P (g–l), and A (m–r) classes on the FIRE database. The first and second columns are original retinal images, and the third to fourth columns are the results of the retinal vessel segmentation and junction detection. The fifth column shows the registration image which marks the area where two images meet as yellow. The last column shows registration result by overlapping images.

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i)

(j)

(k)

(l)

(m)

(n)

(o)

(p)

(q)

(r)