Digital BW, The Print

Roger Howard wrote:
> 
> On Dec 4, 2004, at 10:00 PM, Anthony G. Atkielski wrote:
> 
> 
>>The Wogster writes:
>>
>>
>>>For archiving though, where you are not looking at the image on a
>>>regular basis, you can use other compression mechanisms.  For example 
>>>if
>>>you can compress a 600MB file down to say 50MB for archiving, and then
>>>decompress it again later, who cares?  Even if it takes an hour to
>>>compress/decompress.
>>
>>Agreed.  But what file formats would be in this category?
>>
>>Theoretically, you could compress most image files enormously with no
>>loss, given time and space to do an absolutely optimal compression.
>>I've never heard of a file format that is designed for this purpose,
>>though.
> 
> 
> No - assuming no loss, there are real limits to how far compression can 
> take you. It's closely related to Shannon's Law.


It's more related to the type of data, data like databases and ISAM 
files, tend to contain a lot of empty space or "air", therefore it's not 
uncommon to compress a database to 20%, if you ignore rebuildable 
indexes, and dead chains, then you can reduce it to 5%-10%, it's still 
lossless, because you didn't throw away unrecoverable data.  Text files 
contain lots of repeating information, so they compress fairly well also.

> However, what you say is very true if you allow for a little loss... 
> then yes, we can keep devising deeper, more sophisticated, lossy 
> compressors that take more and more memory and CPU. But if the 
> requirement is lossless, then we can do pretty nearly optimal 
> compression today - even throwing tons more CPU at it with much more 
> complex lossless algorithms you'll never get past Shannon.

That may be true, a bitmap consists of lots of non-repeating data, the 
ideal compression would be something that reduces the amount of 
information that needs to be stored.  Something along the lines of 
vector graphics might work. Something else that might work, is to split 
the information out, for example into R/G/B and see if the repeatable 
data amount goes up.

One thing to remember though, Shannon did his work circa 1948, so far 
nobody has challenged his theory, however many, many theories have been 
though impassable before, until someone proves them wrong.

> 
> Today, the best compressors for really great efficiency on really high 
> rez materials tends to be wavelets... JPEG2000 and MrSid are both 
> popular implementations. The artifacts are much less obnoxious than 
> DCT, and the compression ratios are much better. For really large 
> images, unless you regularly view them at 100% then a wavelet 
> compressor (or even DCT) can be used pretty heavily without losing too 
> much significant detail.
> 
The ideal, is to stay away from the idea of repeated Jpeg Load/Save 
cycles.....

W