Encoder Optimization for B&W video

Just thinking that if D2S knew that it was dealing with B&W source video, the encoding might be optimized by reducing the palette used. Of course, several components would probably be involved. And this would be of more interest to rippers of old flicks. Anyway, if there exists a way to implement this, TIA for the info.

Never thought about it before.... I believe there is a avisynth filter "greyscale()" that may (or may not?) convert to B&W,it may kill any brownish celluloid type colour the original has though. Easy to try in an avisynth script and fiddle with anyway. May be worth a bit of a search of the avisynth forum.

There is indeed such a command which converts to black and white. Certainly for sources such as VHS capture I would see this as very useful. I should imagine from digital sources, the studios would have used every possible optimisation when converting from the analogue source in the first place. They would have to, really, since old B/W movie footage tends to be somewhat flickery and noisy and therefore extremely difficult to compress. As for what combination of avisynth filters one may try to "clean up" such footage ready for MPEG2 encoding, I'm afraid that's rather out of my depth.

My thoughts are not towards converting color to B&W. Rather that old flicks are almost always 1.33 to 1 aspect ratio and if you rip say a 90 min BW flic in 1.33 to 1 and a color flic in 1.66 to 1, the letterbox borders reduce the encoding time substantially. I recently did a 2 hr 40 min BW rip full screen that took 26 hrs on a 2.4GHz system. I don't know the internals of the ecoders, but I doubt that a 64K pallette is needed to represent grayscale through the process. (I use TMPGenc if that's relevent.) If 256 or 1024 gray levels were adequate for a BW movie, then the encoder could reduce the encoding algorithm during multi-passes and then map the reduced palette to the standard output format. OK, I'm done. BTW - I use DVD2SVCD because I am not knowledgable or proficient enough to set up my own scripts for the rip process.

I don't think that was quite Ralphthedog's point - as in I don't think he thought you wanted to convert colour footage to black and white. I think he was suggesting that old film source often has a sepia tint to it. Whereas I myself was suggest analogue captures of black and white footage through a capture card are also likely to be filled with purely artefactual colour information. I think ralphthedog was putting forward the idea that adding the greyscale command to the avisynth script would remove such irrelevant information and serve pure greyscale frames to your encoder. I was suggesting that if you are backing up DVD's, the source itself will be pure monochrome as the studios will have removed such artefacts themselves before encoding. As far as I know, there is no option in TMPGEnc for greyscale encoding - ie an option to only use the luminance value of the input data and disregard the colour information - and in RGB24 colourspace even reducing to greyscale in avisynth is of questionable merit, though dependent on source it could improve quality. In short, therefore, I can see no way of speeding the process up. As for your point about full screen and letterboxed encodes, MPEG2 encodes motion, so the fact that there is no motion in black bars makes them very easy to encode. It is not related for to a lack of need for colour information. Leterboxed encodes are quicker regardless of whether the image is colour or black and white. Sorry not to offer any useful solutions for you, but I hope this helps you understand why I feel this avenue comes to a dead end. Cheers Nick

Hi- If I can add a few things. PackRat seems to think that the fact that it's full-screen B&W means that the encoding should go faster when compared to a full-screen color film. That's simply not true. I'm quite certain that if he had encoded a full-screen 2 hr 40 min color movie, that it would have taken the same 26 hours, all other factors being equal. And yes, encoding full-screen (B&W or Color) will always be slower than encoding widescreen (especially 2.35:1 movies) with black bars added, for the reasons that Nick mentioned. I was suggesting that if you are backing up DVD's, the source itself will be pure monochrome ... Not necessarily, and not even usually, in my experience (which consists of encoding hundreds of B&W DVDs). There are frequently quite a bit of moire patterns or rainbows in the DVD source. These are most often seen in and around text and fine lines or patterns. In addition, the filters being used, especially temporal filters, may add color information of their own. When encoding B&W, it's almost a necessity to add some sort of a greyscale filter at the end of the .avs. You can use as ralphthedog suggested. I prefer to use Donald Graft's which is also built into AviSynth.