Normal video is captured, transmitted and stored in an interlaced format. A complete video frame consists of two fields. The first field or first half of the interlace contains all the even numbered lines, while the second field or half of the interlace contains all the odd numbered lines. As most video cameras open their shutter for each field (rather than once for the two fields), if the image is changing with time the information contained by the two fields represents different times. For PAL systems, there is a 1/50s lag between the two halves of the interlace, whereas for NTSC there is a 1/60s lag; a complete frame is available every 1/25s or 1/30s (respectively).
Now if the video signal is of a rapidly moving or evolving event, then this time delay might represent a significant difference between the even and odd lines. Consider a vertical bar moving rapidly across the viewing area. Suppose that when the camera is scanning line 0 the bar is at column 1. By the time the camera finishes scanning the even lines the bar may be at column 100. The image in the even field will correspond to a slopping bar. By the time the first odd line is scanned the bar may have moved as far as column 120, and will be at column 220 by the end of the odd field. Thus the total video frame will appear to contain two sloping bars, one present on each half of the interlace.
The purpose of the Average and Duplicate filters is to create an image containing the information in only one half of the interlace, thus specifying more exactly the time at which the image is taken. After using this filter on the above example, only one bar (still slanted) will be present in the image. The filter operates by replacing the odd (even) numbered lines with either the average of even (odd) numbered lines to either side, or simply with the line above (below). Both these filters effectively discard half of the information present in the original image; the average filter does so in a manner which produces a smoother resultant image.
A further use of these filters is in some situations where the frame grabber fails to lock onto the video sync signal adequately. This frequently occurs when trying to acquire an image from a VTR in pause mode, but may also occur when acquiring an image from a stretched or worn region of video tape. Under these circumstances the Data Translation frame grabber hardware tends to have difficulty with digiting the first video field, but manages to produce a reasonable image with the second field. A useful image may still be obtained by replacing the even numbered lines by a duplicate or of the neighbouring odd lines.