me | department | university | teaching | research | disclaimer | personal stuff



Registration

Test Sets

The following test series were used to determine the efficiency of the different registration techniques. Each test set contains images of the same specific characteristics. For this WWW survey we chose three representative images of each test set.

Validation

The validation tables first compare the similarity of the first pair of images (BP1). For that purpose we use different similarity measures: Direct Comparison (DC), a comparison based on Bounding Boxes (BB), a comparison based on Mutual respectively Normalized Mutual Information (MI, NMI), and a comparison based on cross-correlation (CC). All these similarity measures are added to the total amount of the validation result. (The higher the total amount of the validation result is, the more accurate is the registration process of the images). The transformation techniques used are rigid and affine PCA-registration (rPCA and aPCA), Monte Carlo Warping (MC), and a combined method (PMC).

sample image sample image sample image sample image sample image

These test series were used to determine the efficiency of the different registration techniques.

Registration results

The following images were produced using different registration techniques. We subdivide the images in two parts (affin and elastic). Image (a) shows the starting image and image (b) shows the target image.

sample image

sample image

affin transformations elastic transformations



[Image of CS Logo]
file last modified: Friday, 26-Feb-2016 12:34:25 CET
email address
Copyright © 2017 Martin Held. All rights reserved.
[Rock climbing]