Inverse Telecine (IVTC) doesn't reduce file size

[johnmeyer]

I am converting films on my laserdisc collection to DVD. I capture via Firewire through my camcorder to NTSC DV at 29.97. I then put this into TMPGEnc.

Here is my question:

If I use TMPGenc's Inverse Telecine feature, and then encode at a frame rate of 23.976, I get the same file size as when I do NOT use the IVTC feature and encode at 29.97. I use the same bitrate and encoding format (CBR) for both. (I tried encoding at 23.976 both with and without the pulldown flag).

I was expecting to see a file size that was 23.976/29.97 = 80% of the original size. I was then going to increase the bitrate to get the file size back up to the same as the 29.97 encode, thus getting not only the quality increase from encoding the intermediate blended frames created by the telecine process, but also by encoding fewer frames.

Why doesn't the file size get smaller when using IVTC??

[Mythos2002]

I'm not sure if the filesize will change. Did you do something like the example below? That is the way I usually do IVTC.

Did you choose 3:2 pulldown when playback under Encode Mode of the Video Tab? If so, you should choose 23.976 fps (internally 29.97 fps) under Framerate. Also, under the Advanced Tab, did you click the Inverse Telecine filter? You really have to double click it and a screen will come up where you can tell it to do something. Your source may vary, but I usually choose Auto Setting and then the Setting Method of Flicker Prioritized (24 fps) and Deinterlace: none.

Hope that helps.

Mythos

[johnmeyer]

Did you choose 3:2 pulldown when playback under Encode Mode of the Video Tab?

Yes. My settings on the Video tab are:

Stream Type: MPEG-2 Video
Size: 720x480
Aspect ratio: 16:9 Display
Frame rate: 23.976 fps (insternally 29.97)
Rate control mode: 2-pass VBR (VBR) (Setting 4750 avg; 8300 max, 1000 min
Bitrate: 7000 kbits/sec
VBV buffer size: 224 KB
Profile & Level: Main Profile
Video format: NTSC
Encode mode: 3:2 pulldown when playback
YUV format: 4:2:0
DC Component precision: 10 bits
Motion search precision: High quality (slow)

Also, under the Advanced Tab, did you click the Inverse Telecine filter?

Yes. The Advanced tab settings are:

Video source type: Interlace
Field order: Bottom field first (field B)
Source aspect ratio: 4:3 525 line (NTSC, 704x480)
Video arrange Method: Full Screen (keep aspect ratio)

The following two items are checked:

Inverse telecine (I have it set to Convert to 24 fps: Flicker prioritized, and in the IVTC dialog, I used the Auto-setting to detect the pulldown automatically and adjust, if necessary, if the pulldown changes. Like you, I did not check deinterlace.

Clip frame is selected, and I have clipped the black on the original (which was letterboxed.

I kept the GOP structure settings the same as the standard NTSC DVD template.

On the Quantize Matrix tab, I checked the "Output YUV data as Basic YCbCr not CCIR601" after I got a washed out encode and realized that, since I had captured using a DV codec, the colors were alread mapped to 16-235.

With these settings, I can do an A/B between the laserdisc and my DVD-RW test disc and I cannot tell the difference on a 45 inch TV screen.

However, I am still puzzled as to why the file size doesn't change when I am supposedly encoding only 80% of the number of frames.

[adam]

The filesize will not change. You set your bitrate per second not per frames. An IVTC reduces the number of frames per second, but its still the same number of seconds and thus at the same bitrate it will still be the same size regardless of whether you leave it at 29.97fps or IVTC to 23.976fps. A 2hr movie will always be the same size when encoded at the same bitrate, regardless of whether the fps is 24 or 100.

Now since you have less frames every second, but the same amount of bitrate every second, you are getting more bits per frames; about 20% more. This will surely increase quality. If you wish you can lower the bitrate, get a smaller filesize, and still keep the same level of quality as you would have gotten using 29.97fps at the original bitrate. THAT is where the filesize difference plays in.

Since you have a defined output size (the size of your dvd-r) there is no reason to worry about size. Just use the best bitrate you can get in that amount of space. Use a bitrate calculator and you will totally fill the disk. And perform the IVTC to increase the quality. In my opinion the quality increase is massive. Its more then just more bits per frame, you are now encoding progressive footage instead of interlaced footage. That is the main benefit.

[johnmeyer]

Adam,

Bingo!

You set your bitrate per second not per frames.

But of course! bits per second. File size is directly proportional to total bits, and total bits equals bits per second times the number of seconds of video. Whether I have ten frames or a million frames makes no difference to file size. However, if I have fewer frames, and keep the total bits and therfore file size the same, each frame will be encoded at a higher quality.

I understand now completely.

... perform the IVTC to increase the quality. In my opinion the quality increase is massive. Its more then just more bits per frame, you are now encoding progressive footage instead of interlaced footage.

I couldn't agree more. I first found out the power of IVTC many years ago when I was trying to make great SVCDs. The difference in quality was staggering. Without IVTC, when the encoder encountered frames that were interlaced by the telecine process -- after seeing several preceding frames that were nice, clean progressive frames -- it tried to figure out the motion estimation vectors for all those herring bone patterns. It just went nuts, and the "mosquito noise" covered the screen.

I just got finished encoding my Star Wars laserdisc to DVD. On my first encode I forgot to adjust for the 16-235 problem (I captured using a DV encoder) and got washed out color. On my final encode everything went perfectly. I queued up the laserdisc and my DVD and did an A/B test on a 45 inch monitor. I cannot tell the difference at all.

Thank you VERY much for your reply.

[Mythos2002]

Adam,

I just got finished encoding my Star Wars laserdisc to DVD. On my first encode I forgot to adjust for the 16-235 problem (I captured using a DV encoder) and got washed out color. On my final encode everything went perfectly. I queued up the laserdisc and my DVD and did an A/B test on a 45 inch monitor. I cannot tell the difference at all.

Thank you VERY much for your reply.

Thanks for the info. Adam. I should have known better myself about the filesize not changing, but I didn't.

John,

I'm doing the same thing. What is the 16-235 problem? I did one encode and had the video go into slow motion for a few seconds during the Greedo/Han scene. Is that the 16-235 problem or does 16 refer to 16:9 and 235 refer to the 2.35:1 aspect ratio? I'm going to redo my encodes for future dual layer burning.

Thanks.

Mythos

[johnmeyer]

I'm doing the same thing. What is the 16-235 problem?

My cryptic reference to the "16-235" problem refers to the way different codecs map intensities. Computer monitors let you display intensities from pure black (0) to pure white (255). NTSC TVs, however, don't like either extreme and the inputs to them are supposed to be limited to 16 as the darkest value and 235 as the brightest. If you capture using the pass through on a DV camcorder, then it takes care of limiting the blacks and whites. If you then use an encoder that takes matters into its own hands and thinks that the video has intensities going from 0 to 255 when in fact they only go from 16 to 235, it will then re-map them again, giving you a pathetic range of only 32 to 215.

The good news is that effect is not subtle, so you don't have to guess as to whether you really have a problem or not. Do a sample encode and if the video looks washed out and has low contrast, then you need to defeat your MPEG encoder's remapping (which some, but not all encoders can do). If you can create a sample DVD on a DVD-RW and play it in an A/B fashion with the original source, the colors and contrast should look absolutely identical if you've done everything correctly. I truly cannot tell any difference at all between my THX letterbox laserdisc of Star Wars and the DVD I made from it (when viewed on a Pioneer 45" rear projection NTSC TV).

I did one encode and had the video go into slow motion for a few seconds during the Greedo/Han scene.

If you did inverse telecine, this could be the result of a change in the pattern used during the telecine process. The advanced encoding schemes (which are often modified by hand) often change the pulldown pattern at scene changes. The better Inverse Telecine algorithms (like the one in TMPGEnc) can usually (but not always) spot these changes and adapt. Various people have claimed that the VirtualDub adaptive Inverse Telecine can also do this, but the documentation for the product states otherwise, so I don't know. I have always had good luck with TMPGEnc's IVTC facility.

[Mythos2002]

I'm doing the same thing. What is the 16-235 problem?

My cryptic reference to the "16-235" problem refers to the way different codecs map intensities. Computer monitors let you display intensities from pure black (0) to pure white (255). NTSC TVs, however, don't like either extreme and the inputs to them are supposed to be limited to 16 as the darkest value and 235 as the brightest. If you capture using the pass through on a DV camcorder, then it takes care of limiting the blacks and whites. If you then use an encoder that takes matters into its own hands and thinks that the video has intensities going from 0 to 255 when in fact they only go from 16 to 235, it will then re-map them again, giving you a pathetic range of only 32 to 215.

The good news is that effect is not subtle, so you don't have to guess as to whether you really have a problem or not. Do a sample encode and if the video looks washed out and has low contrast, then you need to defeat your MPEG encoder's remapping (which some, but not all encoders can do). If you can create a sample DVD on a DVD-RW and play it in an A/B fashion with the original source, the colors and contrast should look absolutely identical if you've done everything correctly. I truly cannot tell any difference at all between my THX letterbox laserdisc of Star Wars and the DVD I made from it (when viewed on a Pioneer 45" rear projection NTSC TV).

I did one encode and had the video go into slow motion for a few seconds during the Greedo/Han scene.

If you did inverse telecine, this could be the result of a change in the pattern used during the telecine process. The advanced encoding schemes (which are often modified by hand) often change the pulldown pattern at scene changes. The better Inverse Telecine algorithms (like the one in TMPGEnc) can usually (but not always) spot these changes and adapt. Various people have claimed that the VirtualDub adaptive Inverse Telecine can also do this, but the documentation for the product states otherwise, so I don't know. I have always had good luck with TMPGEnc's IVTC facility.

Thanks for the information John. I used TMPGEnc's IVTC, but I split each film on two discs. Maybe I messed up somewhere during the editing. ANH was the only one I noticed with the problem. Did you have to do any manual IVTC at all during the process? I know you can tell TMPGEnc to use this setting after this point and so forth.

I was curious about what you said about defeating the MPEG encoder's remapping. Can you do that with TMPNGEnc? If so, how do you do this.

I use an ADVC-100 for capturing so I don't know how well it really handles colors, but I did notice some of my colors on my first encodes were probably not as vibrant or rich as I would like. Did you use any of the simple color correction filters in TMPNGEnc?

Thanks

Mythos

[johnmeyer]

I was curious about what you said about defeating the MPEG encoder's remapping. Can you do that with TMPNGEnc? If so, how do you do this.

The answer was buried in one of my first posts in this thread:

On the Quantize Matrix tab, you need put a check in: "Output YUV data as Basic YCbCr not CCIR601." This will get rid of the washed out colors. Of course if you don't need to do this, you will get too MUCH contrast.

Did you use any of the simple color correction filters in TMPNGEnc?

No, but if you are interested, here is a link to a short tutorial on the subject:

http://www.bealecorner.com/trv900/DVD/tmpgenc-tips.html

[tyee]

Hi John
I just posted in the Vegas forum a message to you about this LD to DVD converison. I'm just starting to do this too, and am wondering if Vegas 5 can properly do IVTC?? If not, I'm going to try decomb with AVISynth.

[johnmeyer]

I tried to find a way to get Vegas to do it, but the 24p telecine feature really isn't set up to do what is needed for getting Hollywood films back to 24 fps. The nasty little secret is that the telecine process (which repeats and weaves frames in order to get 60 fields per second -- 30 FRAMES per second) is that it often CHANGES during the film. Thus, you cannot just assume that some pattern is going to repeat exactly for the entire two hours of the film. Therefore, you need to find a program that has really intelligent IVTC capability (such as TMPGEnc). It uses "intuitive" algorithms to look at differences between adjacent fields, and throw out those that have large differences (indicating that adjacent film frames were weaved together. It also looks for repeated frames (which is also part of the telecine process) and deletes one of the repeated frames.

[tyee]

Thanks for the info John. I'll look into TMPG.