Source and Target Rectangles in Video Renderers

There are three sizes found in the VIDEOINFO, VIDEOINFOHEADER, and VIDEOINFOHEADER2 format structures of video media types. This article explains what they are and how they work.

First, there is a size in the bmiHeader member of these structures. The bmiHeader member is a BITMAPINFOHEADER structure with its own width and height members, bmiHeader.biWidth and bmiHeader.biHeight.

Second, there is a rectangle in the rcSource member of these structures; and last, there is a rectangle in the rcTarget member of these structures.

Assume you have two filters, A and B, and that these filters are connected to each other (A on the left, or upstream, and B on the right, or downstream) with a certain video media type.

The buffers that pass between filters A and B have the size (bmiHeader.biWidth, bmiHeader.biHeight). Filter A should take a portion of its input video determined by rcSource and stretch that video to fill the rcTarget portion of the buffer. The portion of the input video to use is based on how rcSource compares to the (biWidth, biHeight) size of the media type that filters A and B originally connected with. If rcSource is empty, filter A uses its entire input video. If rcTarget is empty, filter A fills the entire output buffer.

For example, assume filter A is receiving video data that is 160 x 120 pixels. Assume also that filter A is connected to filter B with the following media type.

  • (biWidth, biHeight): 320, 240
  • rcSource: (0, 0, 0, 0)
  • rcTarget: (0, 0, 0, 0)

This means that filter A will stretch the video it receives by 2 in both the x and y directions, and fill a 320 x 240 output buffer.

As another example, assume filter A is receiving 160 x 120 video data, and that it is connected to filter B with the following media type.

  • (biWidth, biHeight): 320, 240
  • rcSource: (0, 0, 160, 240)
  • rcTarget: (0, 0, 0, 0)

The rcSource member is relative to the connected buffer size of 320, 240. Because the specified rcSource (0, 0, 160, 240) is the left half of the buffer, filter A will take the left half of its input video, or the (0, 0, 80, 120) portion, and stretch the video to a size of (320, 240) (by 4 in the x direction, and by 2 in the y direction) and filling the 320 x 240 output buffer.

Now assume that filter A calls CBaseAllocator::GetBuffer, and the media sample returned has a media type attached to it, signifying that filter B wants filter A to provide a different size or kind of video than it has previously been providing. Assume the new media type is:

  • (biWidth, biHeight): 640, 480
  • rcSource: (0, 0, 160, 120)
  • rcTarget: (0, 0, 80, 60)

This means that the media sample has a buffer that is 640 x 480 in size. The rcSource member is relative to the original connected media type (320, 240) not to the new media type of (640, 480), so rcSource specifies that the top-left corner (25%) of the input video is to be used. This portion of the input video is placed in the top-left (80, 60) pixels of the 640 x 480 output buffer, as specified by rcTarget of (0, 0, 80, 60). Because filter A is receiving 160 x 120 video, the top-left corner of the input video is an (80, 60) piece, the same size of the output bitmap, and no stretching is required.

Filter A will place no data in the other pixels of the output buffer, and will leave those bits untouched. The rcSource member is bounded by the biWidth and biHeight of the original connected media type between filters A and B, and rcTarget is bounded by the new biWidth and biHeight of the media sample. In the preceding example, rcSource could not go outside the boundaries of (0, 0, 320, 240) and rcTarget could not go outside the boundaries of (0, 0, 640, 480).