What does easy compression tell me about a jpeg?

Discussion in 'Photoshop Tutorials' started by Zak, May 15, 2006.

  1. Zak

    Zak Guest

    Sometimes I compress a jpeg which I have downloaded from the Net.

    I always use the same old version 3.1 of ACDSEE for compression mainly
    because it is simple and convenient although I do have Photoshop and
    Paint Shop Pro. However I tend not to use them for something which
    seems so straightforward.

    As an example, I notice that (with the SAME settings for compression in
    Acdsee) some 500 KB jepgs will compress to 300 KB but other 500 KB jpegs
    will compress right down to a surprisingly small 100 KB.

    What does the final size in KB compared to the original size tell me
    about the properties of original jpeg?

    Or to put it another way, how is the degree of compression (given the
    same settings on Acdsee) linked to the original jpeg's properties such
    as pixel size, print size, etc?
    Zak, May 15, 2006
    1. Advertisements

  2. Nothing. The file size is related to how much detail there is in the image.
    It isn't.

    Andrew Morton, May 15, 2006
    1. Advertisements

  3. Zak

    Mike Russell Guest

    Jpeg uses an adaptive algorithm for compression. Lower quality numbers
    result in smaller files, but the actual size, as you imply, depends on the
    amount of detail contained in the image. File size, and therefore jpeg file
    size does not depend on the pixel size or print size of the image, since
    these are ancillary data that are unrelated to image content.

    For example, a 512x512 image containing only white will save to 14KB at jpeg
    quality 10 in Photoshop, while the same size image containing gaussian noise
    compresses to 695KB. Blurring the noise image by .5 pixel results in a 424
    KB jpeg image.

    Normal photographs will vary between these two extremes, with the biggest
    variable probably being the amount of noise in the image. Certain material,
    such as scanned cartoons, will save smaller and at a higher quality as gif
    images, or RLE encoded tiff images, than as compressed jpegs. Each RGB
    channel is compressed separately, and in theory a somewhat smaller image
    could be achieved by jpeg compressing an Lab image, with a higher quality
    for the Lightness channel.
    Mike Russell, May 15, 2006
  4. Zak

    Mike Russell Guest

    Oops - scratch that. Jpeg images do not compress the RGB channels
    separately, but treat color data separately from luminance.
    Mike Russell, May 15, 2006
  5. Zak

    tacit Guest

    A lot depends on how much the JPEG was compressed to begin with.

    If it was originally compressed with a high quality setting, and you
    save it from ACDSee with a medium quality setting, it may shrink a bit.
    If it was originally saved with an extremely high quality setting, and
    oyu save it with a medium quality setting, it may shrink quite a lot.
    tacit, May 16, 2006
  6. Zak

    Uni Guest

    Sort of, no, more like, how well a text file will zip/compress.


    Uni, May 16, 2006
  7. Zak

    Zak Guest

    So blurring will result in a larger file size? That may explain it. I
    see some very sharp pictures with what seems to be decent detail be
    quite compressible.

    Whereas poorer quality images (perhaps poor scans, images that have been
    tinkered with, poor photo equipment if photo is very old) do not
    compress corrtespeondingly better.

    I thought the poorer the quality of the jpeg the more it would compress
    but I guess that often it is the other way around.

    Is my understanding broadly correct?
    Zak, May 17, 2006
  8. Zak

    Mike Russell Guest

    No - the other way around. Blurring results in a file that compresses more
    Adding noise, sharpening, increasing contrast, all tend to make the
    resulting jpeg larger.
    In general, good clean scans will very little noise will compress better
    than noisy scans. Because most edits result in increased contrast and
    color, manipulated images, with certain exceptions, will generally be larger
    than unmanipulated ones.
    Mike Russell, May 17, 2006
  9. Zak

    2 Guest

    Question Mike: Gaussian blurr is, by definition, not an even blur (although
    it is 'regular'), so adjacencies calculate to differences that do not
    compress well. Is that correct?
    2, May 17, 2006
  10. Zak

    Mike Russell Guest

    Gaussian refers to the "shape" that is used to distribute pixel values when
    calculating the blur. It tapers off gradually, and would produce a slightly
    smaller jpeg file than, for example, a box blur which has sharp edges.
    Mike Russell, May 17, 2006
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.