Mean-shift algorithm for images

Hi,

I want have a program that splits an image up into a number of smaller images.

I now want to cluster these patches to create some prototypes for the input image.

I wanted to do this using a mean-shift but I have NO idea where to start actually coding it.

Could anyone give me some help on this please?

Thanks in advance,

[371 byte] By [neosenatea] at [2007-10-2 9:01:41]

«« JInternalFrame memory problem
»» Multiple return value

# 1

http://www.cs.umd.edu/~yangcj/papers/icip2003.pdf http://herkules.oulu.fi/isbn9514267885/html/a2667.html

MartinHilperta at 2007-7-16 23:08:38 >

top of Java-index,Other Topics,Algorithms...

# 2

Thank you very much for those resources.From experience would you say the mean-shift algorithm is difficult to code up?Thanks again

neosenatea at 2007-7-16 23:08:38 >

# 3

There is a pseudo code in the appendix of http://research.microsoft.com/~cohen/VideoTooningFinal.pdf http://www.ipi.uni-hannover.de/html/aktivitaeten/EARSeL-Workshop2005_Paper/Cellier.PDF http://www.caip.rutgers.edu/riul/research/papers/pdf/mnshft.pdf

MartinHilperta at 2007-7-16 23:08:38 >

# 4

Thanks a lot for those. I've been having real trouble in understanding the algorithm and the pseudo code should help a lot. I tried being lazy and just looking for the code itself but to no avail!Thanks for all your help and resources.N

neosenatea at 2007-7-16 23:08:38 >

# 5

can i have your code, I am developing a module for vehicle tracking system and I am interested in your source code

Chikua at 2007-7-16 23:08:38 >

# 6

Here is the code www.caip.rutgers.edu/riul/research/code.html

chinnikrishnaa at 2007-7-16 23:08:38 >

# 7

hi,

i'm trying to get a deeper understanding of this algorithm for data-driven bandwith selection (for mean shift) http://www.google.com/url?sa=t&ct=res&cd=1&url=http%3A%2F%2Fieeexplore.ieee.org%2Fiel5%2F34%2F26436%2F01177159.pdf&ei=HQc2RtvfEYqEgAS2xrDFDA&usg=AFrqEzehQLZ2zJDCbwGyNGjAPbtIpXycew&sig2=BBcb3prxucuKX_MbWjAfBw

Theorem 1 (Section 3) states that "the bandwith normalized norm of the mean shift vector is maximized when the analysis bandwith H is equal to Sigma". Sigma is the true local variance.

The actual algorithm however (Section 4.1) simply seems to minimizes the distance between the theoretical mean shift (12) and the observed ones (7), leading to (15). is this interpretation correct? In this case, what is the connection with Theorem 1?!

gwenn-ha-dua at 2007-7-16 23:08:38 >

Java programming resources: FAQs, tutorials, compiler and browser download sites, documentation, books lists, IDEs, etc.

Mean-shift algorithm for images

Other Topics New Topic

Other Topics Hot Topic

Other Topics

All Classified