Machine Vision

[chinchilla]

We are using a PCI-1409 to acquire a high def and video rate image from a custom confocal system. We use the BNC breakout box to send in luminance, hsync and vsync from the PhotoMultiplierTube and our custom control box. 

Because of the way we scan using resonators, we have two corrections that we would like to make to the video stream and are unsure whether it can be done at the camera level, i.e., put code into the .icd file we made, or if it is something that needs to be completed in LabView. Any help would be greatly appreciated.

Distortion One: We acquire luminance information in both scanning directions. Our image is thus one where every other line is reversed. Is there a way to reverse every other line and register? Right now we double the effective width of the image and then have to crop the image two times, reverse one image stack and then interleave.

Distortion Two: The velocity of the scanning mirrors varies sinusoidally and thus there is some stretching that occurrs at the edges. How would one go about filtering the stream to correct for this? I assume that this is something that would have to be done in LabView as the pixel clock is fixed.

Attached is the .icd file we modified from defaults to achieve a stable image with the two distortions already mentioned.

-sb

[BruceAmmons]

I don't know any way to do either correction that you are asking for, other than post processing.

1. You could acquire it as a single image, deinterlace the image into separate images, use IMAQ Symmetry to flip the odd lines, then interlace the image back together.

2.  This would require changing the pixel clock during the scan.  Instead, you would need to resample the image using variable spaced pixels.

Bruce

[chinchilla]

Bruce,

Thanks for weighing in on the matter. I suspected as much although I was hoping that there might be a "line-reversal" option through the camera definition file...such an option is available for other video cards that we use.

For previewing and scanning specimens I would like to do real-time...is there an easy way to throw away the odd lines so that I get an image that does not have the pseudo interlacing issue? Then, when I wanted to capture data I could programmatically use another camera definition with the lines and width doubling and then just post-process later?

smb

[BruceAmmons · Accepted solution]

I would just use a single icd file with odd and even lines.  For preview, use either deinterlace or extract to remove the odd lines.  For full images, do the post processing later.

Bruce