Skip to main content
. Author manuscript; available in PMC: 2023 Jan 4.
Published in final edited form as: IEEE Trans Image Process. 2022 Jan 4;31:823–838. doi: 10.1109/TIP.2021.3135708

Fig. 10.

Fig. 10.

Two-step registration results on three examples from the JRC dataset, where source and target images are displayed interlaced as small grids. From top to bottom lines are: (a1) Input images, (b1) Phase-HoG-RANSAC (coarse only), (b2) Phase-HoG-RANSAC + MIND (two-step), (c1) IRR-PWC (input: vessel) (c2) IRR-PWC (input: image), (d1) CoarseNet (coarse only), (d2) CoarseNet + Seg-DeformNet (two-step), (d3) CoarseNet + Phase-DeformNet (two-step). In each example, the leftmost column shows complete images, and the following three columns show magnified details in the yellow, blue and green boxes denoted by numbers. Red Arrows point to misalignment of retinal structures. Red circles show improved alignment results by the second step in a two-step registration pipeline over its global alignment results.