While encoding a series of video clips for Seth Fein yesterday, I noticed that the final output had two large black bars on either side of the frame. Looking back at some of the older clips, it seemed that most of them contained these black bars as well. Not only was this an inefficient method of digitization (black bars eat up file space and degrade overall video quality), but it also seemed to be horizontally “squeezing” the image.
Initially, I thought this might be a problem with the source media (in this case a DVD). If the bars were on the raw video, they could be easily cropped. But no such luck. I then traced the file through each step of the encoding process to determine where the bars were introduced. The standardized process for the Fein videos consists of about 30 steps, during which the raw video is transcoded four times by three separate programs.
1.) The video is captured by EDIUS in raw DV format.
2.) The raw EDIUS file is then “printed” to an .avi (also DV)
3.) The “printed” .avi is loaded into Sound Forge, its audio track is normalized, and then saved.
4.) The normalized .avi is transcoded into separate .wmv and .rm files using ProCoder
The bars were introduced, curiously enough, while saving the normalized file in Sound Forge. The reason for this is complex. But, in essence, it’s because DV format is standardized at 720×480, which is not a true 4:3 aspect ratio if we’re talking about square pixels (for more info, see this website). This is why DV captures sometimes look “stretched.” Sound Forge is confused by, and attempts to fix the ratio, hence the black bars. If the DV file is not normalized, and sent directly through ProCoder, the latter program correctly adjusts the ratio to 4.3 (which was the correct ratio for all of the Fein movies on which I was working).
Here are three clips of John Garfield and Lana Turner to illustrate the different stages of the process.
A screen capture of the DVD:
The slightly elongated DV output from EDIUS:
The black bar version produced by Sound Forge:
So this raises the question – what to do about audio normalization? If unprocessed, the audio track on most of these old movies remains low and uneven. ProCoder has a normalization filter that can be applied during the final stage (thus cutting Sound Forge completely out of the process). Unfortunately, however, the ProCoder filter is terrible. It does not improve sound quality nearly as well as Sound Forge, no matter how it is configured.
So this leaves Sound Forge. I’ve discovered that the black bars problem arises when you “save” the .avi. But if you use the “save as” command and select “Default Template (uncompressed),” it will maintain the original aspect ratio and no black bars will be produced.
Of course, this is far from an ideal solution. I’m still encoding the film twice before the final pass! And there is noticeable quality degradation in the video stream after saving it in Sound Forge. Another alternative would be to extract the audio from the .avi, normalize it separately in Sound Forge, and then re-multiplex the .avi. I need to investigate this further, and look into other options to streamline normalization (especially on macs, if we’re going to be switching from .wmv to Quicktime).
To be continued…