> I need the source code for JPEG2000 .... especially on converting image > to tiles....
JPEG2000 is based upon patented algorithms. Some of the companies involved have not given patent waivers. Other companies have given waivers only for implementing _exactly_ what is in the JPEG2000 standard, which makes the waivers useless for anyone who wants to experiment with the algorithms. If you don't want to experiment with the algorithms, then you don't need the source; recent Matlab releases are able to read and write JPEG2000.
The net effect of this is that no-one can legally give you the source even if we had it, and you _effectively_ cannot (legally) do anything meaningful with the source even if you had it.
If you have a deep interest in JPEG2000, then you will need to buy a copy of the JPEG2000 standard from ISO, and you will need to have a legal team with a lot of experience in international patent law go through the standard in detail; you would then have to negotiate patent rights. This applies *even for research purposes*.
There is an additional serious legal problem: there is a definite possibility that portions of the JPEG2000 algorithms have been patented by companies who have not yet revealed the existence of the patents. It is not simply a matter of negotiating with a known set of patent holders: your legal team would need to protect you against the possibility of lawsuit by additional currently-unknown companies.
Due to these patent issues, I will not take the risk of even reading the JPEG2000 standard, in case I learn something there that might be covered by a presently-unknown patent. I cannot take the risk of accidentally writing code for my work that depends upon something patented and having an unknown company later sue us for patent infringement.
If you want an idea of how messy this can get, research the history of the lawsuits from SCO against some companies which have used Linux: SCO has claimed that patents are infringed somewhere within a fairly large portion of the Linux kernel, but (last time I looked) SCO refused to say exactly which patents were involved or which sections of the kernel were in violation. The estimates I saw when that legal fight started was that it would likely take a minimum of 10 years to settle the lawsuit, at a cost of millions of dollars in legal fees.