JPEG compression options -- can anybody explain?

Discussion in 'Digital Photography' started by Beowulf, Aug 3, 2004.

  1. Beowulf

    Beowulf Guest

    When I save a file as jpeg, I understand basic compression and quality (I
    might save as 90% quality or 10% compression, for example; of course I
    save as TIFF or other lossless format for editing and prints). But there
    are lots of other JPEG compression options I do not understand -- amount
    of "smoothing" (0-100); checkboxes to activate "optimize" and
    "progressive"; "subsampling" choices of: "2x2,1x1,1x1", "2x2,1x1,1x1
    (4:2:2)", "1x1,1x1,1x1",; "DCT Method (speed/quality tradeoff)" choices
    of: Integer, Fast Integer, Floating Point.

    Can anybody explain these option for fine tuning the saving of a JPEG
    image, or direct me to a website that might explain the best such settings
    for JPEG image saving?
    Beowulf, Aug 3, 2004
    1. Advertisements

  2. Gordon Richardson, Aug 3, 2004
    1. Advertisements

  3. Beowulf

    deimos Guest

    Smoothing -- artificially blurs the whole photo such an amount. It can
    help compressibility as JPEG works very well with smooth gradients of
    color; sharp edges tend to work against it.

    Optimize -- does a few reordering and newer optimizations to affect file
    size without affecting quality.

    Progressive -- creates a multi-resolution image that is "progressively"
    downloaded and decoded as to allow a web browser with a slow connection
    to preview part of the graphic. i.e. Goes from no detail and pixelated
    to full resolution as it's downloaded.

    Subsampling -- what blocksize it operates on. 1x1 will give you the
    sharpest image, but may bloat the file size unnecessarily. Use only for
    computer generated and rendered material. 4:2:2 should always be used
    for photos. 2x2 is useful only when either recompressing a previous
    JPEG or when there is little sharp detail at all (like thumbnails or bad

    DCT Method -- Accuracy of the DCT algorithem. Use Floating Point
    always. Speed of the jpeg algorithem is generally a non-issue and you
    shouldn't ever have to use an integer format unless you're creating a
    grpahic for a integer processor only device. Effects image quality
    also. Colors become blocky and inaccurate with lower accuracy.
    deimos, Aug 4, 2004
    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.