Origins of red blur in MiniDV video?

May 24, 2008 11:08 AM

Question for my roommate about red blur in shooting MiniDV video.

Here's his question:

I've noticed in my work shooting MiniDV video that often, areas of intense red will blur at the edges. If I turn down the exposure when shooting, however, this effect can be minimized. Why do certain reds blur like this on MiniDV?

Due diligence: I've heard that this is not specific to my camera or MiniDV, but rather to video in general. Is that the case?

Also, in case it's relevant, I shoot on a Sony TRV900, though I've experienced the red blur problem with other cameras as well.

Some people I've talked to have suggested that it's a CCD issue, that perhaps the "red" light sensor is more sensitive than the other two. Others have suggested that it's a question of how red is specified in PAL. One site I found suggested that it's a YUV issue, since in the YUV color space, one green pixel's matched with half a red pixel and half a blue pixel, leading to noticeable red blurs and less noticeable (but still somewhat apparent) blue blurs. I don't know which of these claims is correct, though.

posted by limeonaire to technology (4 comments total)

Could it be the lens? Chromatic abberation, perhaps? (I'm a still-photo guy, so I may be way off.)
posted by cmiller at 11:33 AM on May 24

I'm guessing it's caused by chroma subsampling. In digital video, the luminance signal is sampled at four times the rate of the chrominance channels. With intense reds, the luminance is low, so you see the quarter resolution chroma data more clearly.

See: http://en.wikipedia.org/wiki/Chroma_subsampling
posted by Neon at 12:38 PM on May 24

Intense reds and yellows are a struggle to represent in YCr'Cb' video space. in MiniDV, this combined with the chroma subsampling causes tremendous reduction in what's captured.

Red is perceived as a brighter color, which is why there is more problems with red (as it gets bright in this color space.) BTW, it's not one green, matched with 1/2 red and blue....

It's four luma pixel with two Cr' pixels (spread vertically over two more), and zero Cb'...and on the next line four luma pixels with zero Cr' pixels (interpolated from the previous line), and two Cb' pixels (handled like Cr' from the prior line.) Most of the green is in the Luma channel, but about 20% is in each of the other chroma channels and mathematically derived.
posted by filmgeek at 1:26 PM on May 24

Do you have a sample you can show us?
posted by Pronoiac at 1:19 AM on May 25