Scan & Edge Detail
Published : 29th November 2003
Wade K. Ramsey, DP writes on Progressive Scan . . .
>...as I understand it, all the pixels capture at the same instant and their >charges are stored. Then the system clocks out the charges in order, >from one side of the screen to the other, line by line, because the CRT >that will ultimately reproduce the image scans that way.
It is my understanding that this is why certain HD CCD cameras accumulate errors in edge detailing on the right side of sharply defined lines. Am I the only one who has noticed this phenomenon when compositing bluescreens?
VFX Supervisor/DP LA
>It is my understanding that this is why certain HD CCD cameras >accumulate errors in edge detailing on the right side of sharply defined >lines.
Interesting idea! I can't speak to the CCD explanation, but I have seen ringing, sometimes with multiple rings, to the right of high contrast areas in several HD cameras. I understand this is an artefact of Gibbs phenomenon.
Sounds like you are compositing using HDCam as the source material. If budget allows, try using D5 on the next job...you might like what you see.
>Sounds like you are compositing using HDCam as the source material.
Already switched, ages ago. Still see the problem off of raw camera feed at times though.
> errors in edge detailing on the right side of sharply defined lines.
>Interesting idea! I can't speak to the CCD explanation, but I have seen >ringing, sometimes with multiple rings, to the right of high contrast areas >in several HD cameras.
Boy, I haven't looked at Gibbs in years (20?). Do real-life edges approach true discontinuous functions though?. But maybe that's not required in order to see the effect. Also, as you know lots of filtering goes on between the CCD's and seeing an image on a display device. There are opportunities for ringing (due to 4:2:2) at many points in the pipe.
Sorry for the quick short-on-content message. Just a quick brain dump 'cause I thought the idea of Gibbs is interesting.
eCinema Systems, Inc.
For those who are curious (both of you), the Gibbs phenomenon or artefact is an overshoot (or ringing) of Fourier and other eigenfunction series occurring at discontinuities. (It can be removed with the Lanczos Sigma Factor and most modern filter design considers this.)
However, true Gibbs artefacts should be symmetrical around the discontinuity (equal pre and post rings, if you will).
This right edge detailing is what I was talking about in the, [cml-HDTV] Playback directly from camera, article stating that it is a degradation due to the playback head of HDCAM. This edge goes away if you avoid the playback head of either F900 camera or F500 deck, when capturing onto a hard drive.
Dhanendra Patel from Sony was witnessing this improvement while we were prepping for my bluescreen shoot, The World of Tomorrow.
The majority of the HDCAM users may not witness this because they don't turn the edge detail and matrix off for shooting, while shooting for composite is a different thing.
Los Angeles, DP