What happened to Wavelet?

It’s no secret that I have a bunch of issues with H.264, so what are the alternatives? I like wavelet, I like it a lot actually, but why?

Source : Wikipedia.org

But let’s take a little detour, and look at delta compression (what I sometimes refer to as removing the temporal redundancy).

Source : DVD-HQ

Now, consider that the biggest problem with H.264 is random access to frames. Instead of referencing the one or more of the previous frames (like any MPEG variant), we always reference the I-frame (first frame in the GOP). That would allow GOPs of very long lengths, minutes if need be. And to decode the last frame we’d only need the first I-frame, and the last P-frame (if there are ANY camera manufacturers that support this, please let me know). As a lot of surveillance video is recording very static scenes, this system would be very efficient in a lot of situations. Not all situations of course, but I am not suggesting an outright ban on MPEG variants, I merely suggesting an alternative.

Now, why wavelet then?

Wavelet has a very neat property. You don’t have to send everything to the client for the client to display the frame. If a client is on a low bandwidth connection, they only get the top left corner of the image. There is no need to transcode or scale anything, you simply skip bytes. There might be situations where a megapixel stream is being shown in a 320×240 window, so why send a giant JPEG across the wire, only to throw away 70% of the pixels?

If we also use the tweaked I/P-frame relation, we can even skip frames if the client cannot keep up, or prefers to get ALL pixels rather than a higher framerate.

Advertisements

2 thoughts on “What happened to Wavelet?

  1. Virgil Kastrup says:

    I work in a completely different industry, but deal with H.264 and wavelet compressed images on a daily basis. Actually wavelet compression is fast becoming a standard in both acquisition and delivery in the entertainment industry.

    ARRI’s alexa camera generates ProRes 4444 files, which are 12bit wavelet compressed quictimes, RED cameras generate r3d files, which are 12bit wavelet compressed files.
    Cineform, which is an industry standard intermediate codec, is wavelet compressed.

    Digital delivery for cinemas across the globe as DCP’s are essentially jpeg2000 wrapped as mxf files.

    For Online editing and vfx work, we need realtime access to any frame at any time.
    We need lossless (or visually lossless) compression and, due to the massive amount of high resolution images constantly piling up on our RAID’s, we need it all in a lightweight codec.

    Needless to say, I’m no big fan of H.264 for anything else than uploading previews to that interweb thingy, but I am(along with thousands of others) forced to deal with it, due to the enormous popularity of Canon 5D mkII. I can’t even guess how many hours we’ve wasted processing and transcoding to other codecs or image sequence formats.

    Now we just need some kind of crossbreed between jpeg2000 and OpenEXR that holds 32bit floating point HDR images and multiple channels of information such as alpha mattes, normals, z-depth and so on, that can be used for realtime playback at high resolution (2K, 4K and above).

  2. prescienta says:

    I was looking into wavelets using GPU acceleration. I did a google search, selected “images” and was horrified. Don’t do it.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: