ushere said, "while i fully appreciate our general aim for perfection, i have to question how many of our end users (ie. general viewers NOT pro's) would even notice the subtle difference between one method and another?"Yeah, over in another thread, I posted a "quick & dirty tutorial" which would probably work fine for most users, most of the time. However, I decided to start this thread with the nastiest clip I could create, so (hopefully) we could finally nail down a "best practice" once-and-for-all.

We'll see. Furthermore, this thread is quickly getting complicated enough that others with a much better knowledge of the subject than I, may be a better judge of what that best practice is.

...Jerry

Furthermore, this thread is quickly getting complicated enough that others with a much better knowledge of the subject than I, may be a better judge of what that best practice is."Best Practices" is always a tough thing to define. This is especially true in this thread.

First, we all need to stick with the exact wording of the original post, namely that we are trying to devise the best way, in general, to make the best looking video on a DVD when starting with HD video. I note that you got into a question in another thread about this, and I don't think the confusion ever got fully resolved. Obviously you can put any file you want on a data DVD, but that isn't relevant. What is implied by the original statement in this thread is that we want to make an NTSC 29.97 DVD that uses the VOB file structure so that it can be played on a consumer DVD player attached to a TV set. That TV set is designed to play interlaced video, whether or not the TV itself displays interlace natively (CRT) or whether it has to do some magic internally to display it correctly.

Now, having said all that, the test case used in this thread creates moiré patterns. If the goal of this exercise is to reduce or remove those patterns, then Bob's (farss) approach of blurring definitely makes some sense, but of course it will decrease the overall sharpness of the resulting video.

But here is the point of my post:

This definitely is NOT a "best practice" that you would want to use on most AVCHD (or HDV) to SD DVD projects.

AFIK, what I posted above is what will work best for most people most of the time, and it involves using the Vegas presets, with the exception of the Sharpen fX.

Finally, Bob misunderstood why I was puzzled by the result of the pre and post toggle. I am well aware that every single time using the Sharpen fX is mentioned in these forums, he always posts that it makes a big difference whether this is applied before or after compositing (the "pre/post toggle"). He is absolutely correct, and having read it many times, I am well aware of that. My only surprise was that it seemed to change the contrast of the clip as well, but I provided an explanation, which I think is what Bob was saying as well, namely that the edge enhancement creates slightly darker and lighter lines around areas where adjacent pixels are at opposite ends of the luma spectrum (bright to dark in the space of just a few pixels). This makes the video look more contrasty and is reflected in a histogram that appears more spread out.

I was also slightly puzzled that I would still see any effect with the Pre setting, but I most certainly did, and if you want video that looks as much like the HD original, rather than artificially enhanced like SD NTSC video normally looks, then this would be the setting I would recommend.

But, in summary, if the goal here was to minimize moiré, then everything I wrote in my first post is completely useless. The moiré problem is unique to this particular clip, and efforts to reduce or eliminate it, while interesting and useful, would create havoc and poor results if applied generally to ALL such projects.

If, on the other hand, you want a workflow that is "correct" for most HD to SD DVD projects, then I think I provided a pretty good starting point -- although I'd love for others to chime in with modifications.

OK, here's my humble offering.
My "cheat" proves that if a free utility can give results like this, Vegas could do a lot better than it is.
http://vimeo.com/17179383
I approached the task from a graphics/print perspective: Reduction + tight pattern = moire, unless there's some damn good dithering going on. Be sure and download the source MPG.

Musicvid beat me to it. I used Microsoft Image Resizer (part of the free Power Toys set) to do the batch resizing. I had problems loading the sequence back into Vegas, but that's another story.

What is the best way to view the mpg files? I don't have a decent interlaced TV any more and watching interlaced footage on the computer gives results dependent on the player and its settings.

"My "cheat" proves that if a free utility can give results like this, Vegas could do a lot better than it is"

It possibly could if you gave it a chance.
Seems to me you've merged the fields which is possibly fine when you're only delivery SD and the motion is such a mess to begin with.

Arguably Vegas could implement a better low pass filter when scaling. The problem seems to be doing that would produce even longer render times. My understanding is that the ultimate solution to preserving the maximum detail while avoiding aliasing requires infinite processing time.

Bob.

Jerry, very interesting and informative thread. I'm still not convinced that shooting and downrezzing 60i produces less moire and jittering than 30p. Any chance you could shoot the same scene with a 1080p30 cam and upload that file for play?

/jerry

No one mentioned using the "Reduce Interlace Flicker" switch in their workflow. I'm curious as to what this switch actually does under the covers. Is it just applying a slight vertical blur, or something more sophisticated?

/jerry

johnmeyer said: "The difference between rendering with the NTSC pan/crop applied at the event level and not doing this only affects whether there are bars in the final output. It did not affect the quality of the output in any other way. Since you don’t want to have bars, I strongly recommend always applying the correct pan/crop to all events. For this clip, the correct pan/crop is Width: 1920 and Height: 1056"John,

I'm going thru your procedure and I'd like some clarification on this point. In the Pan/Crop settings, should the "Source->Stretch to Fill Frame" be set to "Yes"? And the follow-on question is: Under the Render-As Settings, the "Stretch video to fill output frame (do not letterbox)" should NOT be checked - correct?

I guess if I thought this thru, it would be obvious, but right now my mind is becoming boggled {grin}

...Jerry

" Any chance you could shoot the same scene with a 1080p30 cam and upload that file for play?"/jerry,

The best I can do is point you to a 1920x1080 PF30 (30p in a 60i wrapper). Unfortunately, I don't have a camera that shoots natively in 30p. You can get the file here: PF30Capture.zip

Good Luck!
...Jerry

PS: I typed that link from memory, if it doesn't work, I'll fix it in a few minutes.

"No one mentioned using the "Reduce Interlace Flicker" switch in their workflow. I'm curious as to what this switch actually does under the covers. Is it just applying a slight vertical blur, or something more sophisticated?"Over in another thread, I recommended using "Reduce Interlace Flicker" and posted the results where the moire/flicker problem on the subject's herringbone jacket was "fixed" while retaining the sharpness of the clip.

However, I attempted to use it on the clip that is the basis for this thread and it helped a litte - but, alas, the moire/flicker was too severe.

I am interested in hearing others chime in on the second part of your question (i.e. what exactly does "Reduce Interlace Flicker" do?).

...Jerry

A question for JohnMeyer (his workflow summary is listed below) --

I currently shoot HDV in hif-def mode but deliver on SD DVD. I am new to advanced topics such as this. Would you suggest the same basic workflow for 1080i HDV? I realize the pan crop settings might be different, but the rest would still apply?

This is a great thread and it really helps those of us who need a "Monkey See, Monkey Do" approach!

Kimberly

==========

Summary

If I were writing a cookbook for how to get the best NTSC DVD Widescreen video from 1920x1080 AVCHD (and probably 1440x1080 HDV), here is what I would recommend:

1. Use the “match” feature in the Properties window to make sure the project matches your video exactly.

2. In the Project Properties dialog, change “ Full-resolution rendering quality” to “Best,” and set De-interlace method in the project properties to “Interpolate Fields.”

3. Set the Pan/Crop for each event to 1920x1056.

4. Add the Sharpen fX to each event and set the sharpening amount to zero. Set the pre/post toggle to Post if you want a strong sharpening effect that will mimic what “old-fashioned” SD video looks like. Set it to Pre if you want the video to have slightly better resolution (than you would get with no Sharpen fX at all) without the obvious sharpening artifacts.

5. You can do #3 & #4 on the first event, and then use Copy and then Paste Attributes to propagate these settings to all other events on the timeline. Do this before you do anything else or else you will screw up your existing pan/crop and fX settings. A good multi-purpose script program like Excalibur or Ultimate S gives you even more flexibility as to when and how you change these settings.

I have absolutely no idea how IView handled the fields, or if at all; however that had practically nothing to do with getting rid of the moire from downsampling the wavy crosshatch pattern. That is why I suggested Jerry was being a bit diabolical in the first place.

If I downsample to 720x480 in my old PS5, which apparently does not dither much, regardless of the deinterlace method used (or whether one was used at all), then I get this.



In fact leaving the interlace motion edge artifacts in place may actually help to break up the vertical moire a bit.
In any event, Vegas' downsampling seemed to be even worse in this respect.

But I knew from experience that IView does a very good job in this respect; it must have some major dither/demoire built right into its resampling mechanism.



"I approached the task from a graphics/print perspective: Reduction + tight pattern = moire, unless there's some damn good dithering going on."
Moire was a major consideration when I was a college pub assistant and began working with 4-color halftone separations. We often had to register and tape the negs to make a galley proof. Another term for that is masochism.

The situation most people here will be able to relate to is trying to scan a halftone image, say from a magazine or newspaper.
So I initially saw it not as a deinterlace problem, but a universal problem arising from trying to reduce a grid pattern, which once you reach a certain point, causes striking bilateral moire. My feeling is that every Vegas-centric solution offered in this thread (and some are quite creative!) is actually an attempt to work around that universal fact. However, I humbly recuse myself from consideration for Jerry's winning prize, since I admittedly did not play by the rules.

EDIT: I woke up dreaming of a Lanczos resizing filter for Vegas. Anyone care to take this up? As Bob so aptly pointed out, such a thing would be s-l-o-w!

Happy Thanksgiving (US)!

Answers to various questions from the previous threads:

I'm going thru your procedure and I'd like some clarification on this point. In the Pan/Crop settings, should the "Source->Stretch to Fill Frame" be set to "Yes"? And the follow-on question is: Under the Render-As Settings, the "Stretch video to fill output frame (do not letterbox)" should NOT be checked - correct?You can fill in the numbers I provided, or you can temporarily set the project properties to the NTSC DVD Widescreen preset, and then go to the pan/crop dialog, right-click on the video shown in the dialog and select "Match Output Aspect." Remember to then set the project properties back so they match the source footage (AVCHD or HDV) that you have on the timeline.

In Pan/Crop, both Maintain Aspect and Stretch to Fill should be set to Yes.


I currently shoot HDV in hif-def mode but deliver on SD DVD. I am new to advanced topics such as this. Would you suggest the same basic workflow for 1080i HDV? I realize the pan crop settings might be different, but the rest would still apply?Yes, the same workflow will apply, including the pan/crop settings. The reason the same pan/crop works is that, even though HDV is only 1440x1080, it uses non-square pixels, and when you multiply 1440 by 1.333 (the pixel aspect ratio) you then get the square pixel equivalent which is 1920x1080, the same as the square pixel AVCHD format.


I woke up dreaming of a Lanczos resizing filter for Vegas. Anyone care to take this up? As Bob so aptly pointed out, such a thing would be s-l-o-w!My AVISynth approach did indeed use Lanczos to do the scaling. It did provide better resolution, as measured by the "resolution chart" that is created in the test clip by the portion of cloth that extends from the doll's feet down to the lower right of the video. It did nothing, however, to reduce moire. These patterns are an incredibly difficult thing to deal with, even without the issue of scaling and, as I pointed out in my earlier posts, are at some level simply a matter of physics (interference patters, like you created in high school using a wave tank machine) when you try to create thin intersecting lines using a display made up of either scan lines or rows & columns of pixels. Each type of display has different issues, but both will display moire under the right circumstances, such as this test clip.

So, my point is that you shouldn't kill too much time yearning for Lanczos to do the scaling because the difference will be marginal for most HD to SD tasks. This statement is based on my actual experience serving video from Vegas into AVISynth and using the tools available there to scale, and then serving this into the slightly more capable MainConcept MPEG-2 encoder that is available as a separate purchase directly from MC.

@johnmeyer: " It did nothing, however, to reduce moire. These patterns are an incredibly difficult thing to deal with, even without the issue of scaling and, as I pointed out in my earlier posts, are at some level simply a matter of physics (interference patters, like you created in high school using a wave tank machine) when you try to create thin intersecting lines using a display made up of either scan lines or rows & columns of pixels. Each type of display has different issues, but both will display moire under the right circumstances, such as this test clip."Did you get a chance to view the submission by "dan" above? To my old eyes, it appears that his technique has solved the moire problem while maintaining pretty damn good image quality (although I'm not sure the image quality of the hula dancer is as good as your efforts). Of course he did not follow the rules of the "Challenge", in that he used software external to Vegas, but I think this shows that the problem is, indeed, solvable (although possibly not using the current version of Vegas, solely). I'm duplicating the link to his submission here:



@dan: Could you provide a download link to your 720x480 interlaced .mpg, so we can get a better picture of figurine image quality?

@all: Really good discussion here - I'm learning a lot.

...Jerry

PS: Perroneford also made a valiant (although somewhat complicated) effort at getting the moire under control while maintaining good figurine image quality. IMHO, dan's looked better, but my eyes are old (and the difference may be in the quality of the progressive web version).

Me again. A few more questions on the HD to SD workflow. I hope I didn't miss the answers in the above thread. I hate to pose a question if the answer is right in front of me . . .

Point 3, Pan/Crop.
I don't understand why we need to pan/crop if we are using Match Media Properties and selecting fill frame in our render -- in my case it's HDV. I'm not saying this isn't a good idea; I am definitely an amateur here and I'm just trying to understand.

Point 4, Sharpen
I have Vegas Pro 10. I have a Sony Sharpen FX, but I don't see a toggle for Pre- or Post. I looked around in the options to customize the render, but I didn't see it there. I may have missed it?

Point 5, Copy/Paste the Sharpen FX
I am familiar with the copy/paste on attributes (I love it). Here too I don't understand why we cannot apply the Sharpen at the track level. If we do that, does it clobber everything else?

Thanks so much for this great thread and for your patience answering questions.

Kimberly

========

Summary

If I were writing a cookbook for how to get the best NTSC DVD Widescreen video from 1920x1080 AVCHD (and probably 1440x1080 HDV), here is what I would recommend:

1. Use the “match” feature in the Properties window to make sure the project matches your video exactly.

2. In the Project Properties dialog, change “ Full-resolution rendering quality” to “Best,” and set De-interlace method in the project properties to “Interpolate Fields.”

3. Set the Pan/Crop for each event to 1920x1056.

4. Add the Sharpen fX to each event and set the sharpening amount to zero. Set the pre/post toggle to Post if you want a strong sharpening effect that will mimic what “old-fashioned” SD video looks like. Set it to Pre if you want the video to have slightly better resolution (than you would get with no Sharpen fX at all) without the obvious sharpening artifacts.

5. You can do #3 & #4 on the first event, and then use Copy and then Paste Attributes to propagate these settings to all other events on the timeline. Do this before you do anything else or else you will screw up your existing pan/crop and fX settings. A good multi-purpose script program like Excalibur or Ultimate S gives you even more flexibility as to when and how you change these settings.

"It did nothing, however, to reduce moire."

Actually,my Vimeo link and the second still were resampled with Lanczos, as well as a couple of other free Lanczos utilities I tried with almost identical results.

What the switches are to replicate this ability in AVISynth I have no idea, but I'm certain it's possible.

"PS: Perroneford also made a valiant (although somewhat complicated) effort at getting the moire under control while maintaining good figurine image quality. IMHO, dan's looked better, but my eyes are old (and the difference may be in the quality of the progressive web version). "

Well, I got similar results in Squeeze in one step, but you said we had to do it in Vegas. I simply worked inside the parameters given.

Yes, I could have rescaled with Lanczos or a Bicubic Spline, or Mitchell, but that's not available in Vegas. I usually do my rescales in Virtualdub (or used to) where I had control and good options.

Did you get a chance to view the submission by "dan" above? To my old eyes, it appears that his technique has solved the moire problem while maintaining pretty damn good image qualityYes I did. Dan, of course, will forget more about video in the next five minutes than I will ever know in my lifetime. However, one thing I should point out:

he uploaded 720p video to YouTube.

What this means is that the video is not 720x480 but instead is 1280x720. The less the video is subsampled, the fewer problems you'll have with the artifacts we've been talking about. As another poster said, I'd like to see the 720x480 MPEG-2 file rather than something that has been changed to a high resolution, deinterlaced and changed to progressive, uploaded to YouTube, resampled by YouTube, and then displayed on my computer screen. All those extra steps create a completely different workflow than what was initially asked for in the original post.


I don't understand why we need to pan/crop if we are using Match Media Properties and selecting fill frame in our renderIf I understand your question, it sounds like there is a new feature in the versions of Vegas after 8.0c that let you fill the frame during render. I didn't know that. It will probably do the same thing as my workflow, but I would recommend testing it to make sure.

On another note (you knew this one was coming, Jerry), here is the same thing rendered to 480p in Handbrake.

Lanczos + Yadif + CQ in HB is a great combination. Renders really fast too. Be sure to download the source MP4.
http://vimeo.com/17205229

Ok,

Since the discussion has morphed into the efficacy of various other tools and methods to produce quality files, I decided to use my current workflow to see how it did. Although Avid has tools designed specifically for scaling video up or down, I chose not to use it, and to do all work in Sorenson Squeeze.

Neither Squeeze nor Avid support MTS files, so I did the initial conversion in Vegas and created a DNxHD 220x 60i file. I took that file into Squeeze, selected Mpeg2 16x9, and set VBR of 9.5k/9.2k/9.0k for my Max/Avg/Min respectively. That process created this file:

http://exposureroom.com/543a3321d6194204a58cf7dc19c22b3c/
(you can view it online, but download it see the actual quality)

I then took that mpeg2 file back into Squeeze and did a multi-pass upconvert to 720p mpeg4 at 6Mbps.

That resulting file is here:
http://exposureroom.com/cf870edadbe24d90948bf37435241767/

Again, for accurate viewing please download the original file.

For those not accustomed to Exposureroom, there is a small down-pointing arrow under the video. Click that to download.

There are some who question why anyone would pay $600 or more for a compression program. The answer is that in Squeeze, I did not add any filters of any kind. No blurs, no sharpening, no deinterlace, nothing. It has the power to batch encode to whatever you want. A single file to many types (so if you need a BluRay, a DVD, and a 2 web versions, you can do that in one go). Or if you've got a dozen different source files and the EACH need various types of encodes, you can do that also. The program can email you when your encodes are complete, and it can automatically publish the files to an online server and email you and your client/producer for client review.

Yep, it costs real money. But when you can get pro level results with no fiddling, no filters, just drag, drop, and encode, the time saved more than pays for itself. For home use, this is complete overkill. But if your job has you doing encodes every day, these types of programs are a very worthwhile investment in my view.

-P

The Lanczos algorithm keeps coming up in these discussions. For the life of me I cannot find anything definative that shows it performs any better than the precise bicubic that Vegas uses. The following quote from Wikipaedia's entry on Lanzcos is interesting:

"Blinn characterizes himself and others who prefer such filters as "negative lobists" for their preference for such filters more optimized for frequency response than for ringing.[4]"

By comparison Bicubic is slower but produces less artifacts. In other words a Lanczos resize will look sharper but doesn't seem to be of any higher resolution.

Bob.

Bob, which bicubic are you referring to?

Bicubic SPLINE is the one that is comparable to Lanczos and can offer superior results. The Bicubic in Vegas is not spline (nor is the one in Virtualdub) and uses methods similar to Lanczos, but it generally not as highly regarded.

In general, I have found Lanczos to be extremely effective in nearly every video application I've needed it, and have found it vastly superior to what I could get out of Vegas. At this point, I've used it hundreds of times on video ranging from our difficult test subject here, to underwater HD video I was preparing for the web, to film shorts, you name it.

Lanczos scaling is generally a top choice of UpRes as well, though we don't see that use nearly as often. However, in some of the work I am doing now, where I have to combine SD and HD sources into an HD program, it's again proving it's worth to me.

I realize not everyone will agree, and that's fine. I am trusting my eyes and the eyes of my clients for my views. Things may be different on a set of scopes.