> Correct me if I am wrong, but JPEG is all about
> "Lossy" compression - the lower the compression
> quality, the higher the loss of information in the
> picture (leading to artifacts, etc.). Now at 100%, the
> loss may be minimal but there should still be some
> compression

Sorry Harmanm, going to have to correct you here. JPEG is all about
lossy, but 100% quality does retain all the detail in the image. I
don't know how much people here know about Fourier transforms but
basically a JPEG stores your image as a bunch of sine and cosine
waves. The fewer waves the better the compression but the more
approximate the stored image (hence the loss). At 100% the number of
sine waves used to store the image is however many it takes to
reproduce the original exactly. Typically this is ~4000 waves (90%
however will need only ~30 waves and produce something almost
indistiguishable). Even at 100% though, this alternative form of
storage can prove to be more efficient (for photographs anyway), and
so the file still will shrink, even  though no information is lost.

JPEG falls down on subjects that don't lend themselves to storage as
waveforms, such a diagrams with lots of hard edges (high frequency
waves). If you were to produce a JPEG at 100% quality for (say) a
checkerboard you might find that the image ends up very large
(although still probably not as large as the original BMP).

Why do BMPs almost always turn out large? Because all the images we
tend to care about are (to a certain extent) predictable. Compression
algorithms work because they use less space on your hard disk to store
the 'predictable bits', since they are 'easy'. BMP uses the same
amount of space regardless of whether an image is predictable or not.
For example, an all white image is really predictable (I've had 60000
white pixels so far, there's a good chance the next is too), but BMP
would still just give the next pixel 8,16 or 24 bits as appropriate
for the file. However, for JPEG, an all white image is a flat wave (ie
a constant, very low frequency) so only one waveform is required to
store it, and it compresses very well.

Hope this helps

David

PS To try and lend credibility to this, I am a software engineer with
a background in compression algorithms and data storage.
*
****
*******
***********************************************************
*  For list instructions, including unsubscribe, see:
*    http://www.a1.nl/phomepag/markerink/eos_list.htm
***********************************************************

Reply via email to