Hi,

I am using some ultrasound analysis methods that pretty much rely on equal time shifts from one frame to the next. The more I dig into this, I believe that the time shifts between frames is NOT constant. I hope that someone here can help me explain what I'm seeing.

To understand this better, I recorded some videos where I added a 100 hz wave signal to the video (green line in the bottom of the frame).

The original signal that the ultrasound machine creates is 101 hz. This is being stored in the ultrasound machine as 59 hz, and it is being exported to a PC through an S-VHS port, i.e. 25 hz interlaced.

I have three videos:
The video stored in the ultrasound machine (16 mB):
http://dl.dropbox.com/u/10671621/trigger_Philips_original.avi

The video stored in the PC, when the PC is set to NOT deinterlace (512 kB):
http://dl.dropbox.com/u/10671621/trigger_PC_deinterlace_off.avi

The video stored in the PC, when the PC is set to DO deinterlacing (2,3 mB):
http://dl.dropbox.com/u/10671621/trigger_PC_deinterlace_on.avi

I expect that my 100 hz wave signal moves through the video frame at a constant speed. Obviously the frame rate and the movement speed of the wave do not match up. Then I would expect that the wave signal moves in a symmetric fashion, i.e. 1 pixel from frame 1 to 2, 2 pixels from frame 2 to 3, then 1 pixel, then 2 pixels, etc. Or 1112111211121112 etc.

Looking at the video from the ultrasound device, the signal does not move in a symmetric fashion - mostly it moves 1 or 2 pixels, but sometimes 3 pixels, sometimes 4 pixels. There is no system - the wave moves this many pixels within the first twenty frames: 21122122122122122222

So, is the downsampling done in a more or less random fashion?

The situation gets further complicated when looking at the PC file. In many parts of the clip, there is no movement between two subsequent frames (just as if the video wasn't interlaced in the first place). In some parts of the clip, there is movement between each of four subsequent frames. I.e. very little movement between frame 35 and 38, and then large movement between frame 38 and 40.

Is this standard behaviour for video devices that need to change the sampling frequency?