CPC G06T 7/337 (2017.01) [G06T 3/02 (2024.01); G06T 7/37 (2017.01); G06V 10/761 (2022.01); G06T 2207/10081 (2013.01); G06T 2207/10088 (2013.01); G06T 2207/20221 (2013.01); G06T 2207/30016 (2013.01)] | 16 Claims |
1. A two-stage medical image registration method based on progressive images (PIs) comprising the following steps:
step 1: merging a reference image R to be registered with a floating image F to be registered to generate multiple intermediate PIs;
step 2: registering, by a speeded-up robust features (SURF) algorithm and an affine transformation, the floating image F with the intermediate PIs to acquire coarse registration results;
step 3: registering, by the SURF algorithm and the affine transformation, the reference image R with the coarse registration results to acquire fine registration results; and
step 4: comparing the fine registration results of the intermediate PIs acquired in repeating steps 2 and 3 and selecting an optimal registration result as a final registration image,
wherein the multiple intermediate PIs are generated as follows: taking the intermediate progressive image as a reference image for an intermediate registration process and merging the reference image with the floating image F to generate the multiple intermediate PIs; or, taking the intermediate progressive image as a floating image for the intermediate registration process and merging the floating image with the reference image R to generate the multiple intermediate PIs, and
wherein in step 2, the coarse registration result is acquired as follows:
sub-step 1: taking an intermediate progressive image Mk as a reference image and extracting, by the SURF algorithm, feature points of the floating image F and the intermediate progressive image Mk, k=0, 1, 2 . . . l;
sub-step 2: describing the feature points to acquire feature vectors and performing feature point matching based on a similarity of the feature vectors; and
sub-step 3: spatially transforming, by the affine transformation, the floating image F, and calculating, by a least squares (LS) method, a transformation parameter; transforming the floating image F into a coordinate system of the reference image R through the transformation parameter; and registering, by a bicubic interpolation function, the transformed floating image F with the reference image R to acquire an initial registration result described as the coarse registration result.
|