Q on Noiseware and compression

[photog1107]

I use Noiseware plugin for PSE on a Mac. When I save files after doing work in PSE I generally choose either 100% or 80% compression for my final saved file. Usually my files are very close to the original file size. But when I use Noiseware (which is a great product for noise reduction) and then save, my file sizes are drastically reduced (more like 25% of original size. What is going on? Does Noiseware take out that much info from the photo? They look fine for screen/web viewing, but I am worried about what this might do when I go to print. Any info on this from Noiseware users or moderators would be appreciated!

[kazuo]

Hi,

I'm a lurker on the forums (as I don't have a camera system yet =[, only enjoying reading photography books for now (well.. I'm theoretical physicist in the end, so I'm doing theoretical photography for now))

But, as nobody answered you, and is a field that I like to study (compression algorithms), I registered to post a reply.

What is happening is the usual, and perfect normal. When you use an image format like JPEG it use a mathematical algorithm for compressing, reducing the disk space used to represent the data (your photo). JPEG, and practically all the compression algorithms, use repeated pattern for compressing. For example look at this sequence of characters:

aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa

Its have 40 chars (think that 1 char use 1 kilobyte of space, so the number of chars is the "file size"), its big, but it can be compressed as

40 times a

and only use 10 chars, a reduction in size of 4x but its have the same information.

Noise, is a random data that get added to your picture, and random data is something with very little (or none at all) repeated pattern. For example, if we add a random noise in the previous example, like this

aaaaaaaaaaaaaaaaaaaaBaaaaaaaaaaaaaaaaaaa

Now our compressed data is:

20 times a plus 1 time B plus 19 times a

this have 40 chars, no compression at all! The noise (the B char) broke the simple pattern we had, and made harder to find a long repeated pattern.

JPEG format use an algorithm like this one, but with a lot more complications (chroma subsampling, discrete cosine transform, Huffman coding... and is lossy) but the same reasoning works, noise (random data) breaks patterns and decrease the effectiveness of the compression.

But, your question "Does Noiseware take out that much info from the photo?" Yes and no. Noiseware remove info, but as all (good) noise removing algorithm its try to remove an info that matter less to your eyes/perception.

If you eyes says its ok, better than with noise, go for it. In the end photography is for human not for machines. Your file is smaller because computers don't understand that the noise is not (a good) part of the photograph.

In the "40 character sequence" example, removing the noise B don't matter too much "visually", if you don't look close the B noise is not perceptive at all, but the change in "file size" is big.

I'm sorry if I'm not much clear, but be free to ask for a clarification. And sorry for my english.

PS. If you like to understand a little further, the big point is 'random data have more information than organized data', so a chaotic room have much more to say than an organized one. But this is the field of "Information Theory" and big names like Shannon, Kolmogorov... And a very beautiful research area IMHO.

[photog1107]

WOW. That reply blew me away. Thank you for registering on the forums to provide so much insight into such a complex area! I certainly followed the logic of what you explained and as you said, in the end, photography is a visual art, so it's what your eye sees that really matters. I just wanted to make sure I wasn't doing something wrong in the process that would end up damaging my files.

Thanks again, and I hope you get your camera gear sometime soon, so you can join in the fun behind the lens, too.