Talk:Image registration

From Wikipedia, the free encyclopedia

"A basic image similarity-based method consists of a transformation model, which is applied to reference image coordinates to locate their corresponding coordinates in the target image space"

-> i thougt the transformation is applied to the target image and the reference image is unchanged ...

Reply: Although there are some individuals/systems who take the approach of applying transformations to the target image, I believe that it is generally preferred to do so to the reference image. Transformations are lossy (there is information lost in the process) and reference images can often have an order of magnitude more information given their significantly higher resolution. Therefore it is preferable to not further degrade the target image information by applying a transformation to it but rather to the reference image which may result in a transformation path that does not degrade the information in the target image at all. [Disclaimer: I am just starting to investigate the algorithms involved so this information may be inaccurate - please correct me as needed.] Gfariello 19:37, 18 February 2007 (UTC)

[edit] Similarity-Based Methods

The author gives cross-correlation and mutual information as examples of similarity-based methods. If I recall correctly, these examples are also sometimes referred to as template matching registration algorithms. Is this the case, or is my memory muddled? Lemasterda 19:35, 3 October 2007 (UTC)