Inside Macintosh: QuickTime Components
QuickTime handles images in bands, which are horizontal strips of an image. Bands allow large images to be accommodated even if the entire image cannot fit into memory. The Image Compression Manager calls the image compressor component once for each band as the image is compressed or decompressed.
The Image Compression Manager determines the height of a band based on the amount of available memory and the bandMin and bandInc parameters provided by the compressor component in the compressor capability structure (described in "The Compressor Capability Structure," ). The bandMin field specifies the minimum band height supported by a decompressor component. By providing a minimum height, decompressor components that operate on blocks of pixels can operate more efficiently since the minimum height ensures that a band has at least one row of pixel blocks. The bandInc field specifies the increment in pixels by which the height of a band is increased above the minimum when sufficient memory is available. This specification allows easier processing by ensuring that a band is an integral number of rows of blocks. The larger these two parameters, the more memory is required for the band buffer, which may limit the size of images used with a given amount of memory. By specifying a minimum height that is the size of the image, the compressor component can indicate that it cannot handle banded images. However, the specification of a full size is not recommended unless required by the compression format, since it requires large amounts of memory for large images.
For decompressing sequences of images with temporal compression, the Image Compression Manager always allocates the band to include the full image. The entire image must be available whenever the screen needs updating and the current frame does not have information for all pixels. The entire image is needed to make the comparison with the previous frame.
The depth of the band is determined by the Image Compression Manager and the wantedPixelSize field of the compressor capability structure (described on The Compressor Capability Structure ). That field is filled in by the image compressor component's CDPreCompress or CDPreDecompress function (described on CDPreCompress and CDPreDecompress , respectively). The Image Compression Manager requests the depth that it decides is best for the image, and the compressor component can return the wantedPixelSize field set to that depth or another appropriate depth if the compressor cannot handle the one requested.
The width of the band is usually the width of the image, but the compressor can extend the measurement if it cannot easily handle partial blocks of pixels at the edge of the image. For compression operations, the Image Compression Manager sets the extra pixels added to the right edge of the band to the same value as the last pixel in each scan line. For decompression operations, the Image Compression Manager ignores the pixels that were added to the right edge for the extension.
Image compressor components can also use extension for the height of the last (or the only) band in the image (the other bands should always be an integral multiple of the bandInc field set by the decompressor component). The extended pixels are added to the bottom of the band. For compression operations, the added pixels have the same value as the pixel at the same location in the last scan line of the image. For decompression operations, the added pixels are ignored. If an image compressor component does not want to deal with partial blocks of pixels, either horizontally or vertically, it can use this extension technique. However, it would be more efficient for the compressor to handle those blocks itself.