No absolute answer possible Depends on what you intend to mean by a small or
large distance between images. If you wish to measure distances as human eye
might see them then you might want to transform the intensities so that
there is less difference between levels of gray that humans have difficulty
distinguishing (darker shades of gray). 

 

----------------------

F. James Rohlf, Distinguished Professor

Dept. Ecology and Evolution, Stony Brook University, NY 11794-5245

 

From: Classification, clustering, and phylogeny estimation
[mailto:[log in to unmask]] On Behalf Of Yakir Gagnon
Sent: Friday, May 14, 2010 11:35 AM
To: [log in to unmask]
Subject: A good measure of distance for pixel intensities

 

Hi!
Background: I want to compare two images that are identical in many respects
(pixel-wise) but are different in the pixel intensity values they have. I'm
aware of the many image comparison methods out there, but I want to keep it
very very simple  (for all the image analysists: these two images are
artificial and made through the same process with some differences without
any spatial translation). 

What I am doing right now: is calculating the euclidian distance between the
pixel intensities of the 2 images. The mean of all the distances (if the
image is 500*500 pixels then there are 500^2 distances) gives me a measure
of how similar those images are. 

My question is: is the euclidian distance really the best option when
comparing natural numbers that can only range between 0 and 255, or should I
use some other measure of distance or transform the pixel intensities first?

 

Thanks in advance!

  

Yakir L. Gagnon, PhD student
The Lund Vision Group
Tel  +46 (046) 222 93 40
Cell +46 (073) 753 63 54
Fax +46 (046) 222 44 25
http://www.lu.se/o.o.i.s/7758
http://www.google.com/profiles/12.yakir

---------------------------------------------- CLASS-L list. Instructions:
http://www.classification-society.org/csna/lists.html#class-l 


----------------------------------------------
CLASS-L list.
Instructions: http://www.classification-society.org/csna/lists.html#class-l