Rendering Uncompressed - Still looks worse than original..

[Track]

What I was asking was what your intended destination was. Let's just say that your vid ended up on youtube or some other similar showcase. Can you be sure that it would not be converted to normal ie 30fps playback which would ruin its structure.

I'm not sure. This whole differing frame-rate thing makes my head hurt.

EDIT: I figured out that I need the DirectShow Input Driver to make VirtualDub accept .MKV files, but I can't seem to install it.
I created a plugin32 folder and copied the plugin into it, but nothing has changed.

EDIT2: Okay, I got the plugin into VirtualDub and now it lists .MKV as an open-able file.
But when I tried to open an .MKV file encoded with XviD4PSP, it gives me this error:
"File does not have a video stream."

[jagabo]

VirtualDub for some reason does not support the files output from XviD4PSP..

There are source plugins for most containers (MP4, MKV, TS, MPG, etc.). See the VirtualDub page here for the plugins.

You can use the "files of type" pulldown on the Open File dialog to specify which source plugin to use.

Regarding frame rates, what is your final output? AVI file? DVD? Youtube?

[theewizard]

MJPEG with HUFFY encoding

[Cornucopia]

Some thoughts,

The problem as I see it is this:

You do NOT have uncompressed still pictures, you have already JPEG-compressed (aka LOSSY) pictures). However, if you can live with them having already lost something, it's a moot point.

Your pictures are, as you said, 18MP. That is roughly 5196 x 3464. Roughly 6 TIMES the size of a standard HiDef video (at 1920 x1080). Unless you were doing a "5K" or "6K" video, you would inevitably HAVE to be resizing down to this common display size. Well, in the resizing you will be losing quality, regardless of whatever process you want to follow or tool you intend to use. Or do you intend to be cropping (aka Pan & Scan)?

Your whole "framerate" thing should be determined by 2 needs: How fast do you want the time-lapse to transition from picture-to-picture? and How long do you want the whole thing to last?

Before continuing, I would suggest you answer these and the previous questions, and also post MediaInfo text outputs of a sample of your input pictures and also of your current video output attempts. I do believe we're losing some valuable info in the translation...

And, how many picture files are actually making up your 200MB folder?

Scott

[aedipuss]

average canon full size jpeg is 6MB, so he has maybe 30-40 pictures. uncompressed raw are about ~ 23MB. and you were very close they are 5184x3456.

[jagabo]

How fast the "action" is will also determine what kind of frame rate you want. Pictures of pedestrians walking down the street taken at 1 second intervals may look fine played back at 30 fps. But if the pictures were taken a 1 minute intervals playing them back at 30 fps wont work. On the other hand, pictures of a sunset taken at 1 minute intervals will look fine when played back at 30 fps.

[Track]

VirtualDub for some reason does not support the files output from XviD4PSP..

There are source plugins for most containers (MP4, MKV, TS, MPG, etc.). See the VirtualDub page here for the plugins.

You can use the "files of type" pulldown on the Open File dialog to specify which source plugin to use.

Regarding frame rates, what is your final output? AVI file? DVD? Youtube?

Yes, I installed the plugin and .MKV does show up. However, I still get the aforementioned error and cannot open said file.

Well, I intend to keep the .AVI file that VirtualDub created with Lagarith as a "negative", but I'm going to encode it to .MKV, which is why I need VirtualDub to read .MKV files.

Unless there is another program that can take ACCURATE screen captures, unlike Media Player Classic (anyone?)

You do NOT have uncompressed still pictures, you have already JPEG-compressed (aka LOSSY) pictures). However, if you can live with them having already lost something, it's a moot point.

I don't really see the difference between JPEG and RAW in terms of quality. I see RAW as more of a film negative - makes it easier to change white balance and the like.

Besides, you don't need 100% image quality for TimeLapse. If you did, I'd use a 5DMIII.

Most of the time, I use my super-wide-angle 8mm lens, which is not known for its sharpness.

Your pictures are, as you said, 18MP. That is roughly 5196 x 3464. Roughly 6 TIMES the size of a standard HiDef video (at 1920 x1080). Unless you were doing a "5K" or "6K" video, you would inevitably HAVE to be resizing down to this common display size. Well, in the resizing you will be losing quality, regardless of whatever process you want to follow or tool you intend to use. Or do you intend to be cropping (aka Pan & Scan)?

Yes, I intend to render them at full size, or roughly 5k.

Unless, like I said, the photos are not sharp enough, in which case, I'd go down to 4k or maybe even 1080p.

I've rendered a 400-photo timelapse that IS sharp and it has marvelous detail at 5k.

Of course, this is while viewing it on my 2560x1600 monitor.

Your whole "framerate" thing should be determined by 2 needs: How fast do you want the time-lapse to transition from picture-to-picture? and How long do you want the whole thing to last?

Those are secondary. The main point is how fluid I want the motion to be.

If I render it at 15fps and speed it up it won't have the same effect as if I render it at 60fps but slow it down.

I'll have to perform a bunch of tests once I get this whole encoding deal over with..

Before continuing, I would suggest you answer these and the previous questions, and also post MediaInfo text outputs of a sample of your input pictures and also of your current video output attempts. I do believe we're losing some valuable info in the translation...

General
Complete name                            : C:\Users\Track\Videos\IMG_3339.avi
Format                                   : AVI
Format/Info                              : Audio Video Interleave
Format profile                           : OpenDML
File size                                : 2.12 GiB
Duration                                 : 32s 700ms
Overall bit rate                         : 557 Mbps
Writing library                          : VirtualDub build 32842/release

Video
ID                                       : 0
Format                                   : Lagarith
Codec ID                                 : LAGS
Duration                                 : 32s 700ms
Bit rate                                 : 557 Mbps
Width                                    : 5 184 pixels
Height                                   : 3 456 pixels
Display aspect ratio                     : 3:2
Frame rate                               : 10.000 fps
Color space                              : YUV
Chroma subsampling                       : 4:2:0
Bit depth                                : 8 bits
Bits/(Pixel*Frame)                       : 3.110
Stream size                              : 2.12 GiB (100%)

This is for the video. How do I save media info for photos, again?

And, how many picture files are actually making up your 200MB folder?

46. But that's my smallest folder.

How fast the "action" is will also determine what kind of frame rate you want. Pictures of pedestrians walking down the street taken at 1 second intervals may look fine played back at 30 fps. But if the pictures were taken a 1 minute intervals playing them back at 30 fps wont work. On the other hand, pictures of a sunset taken at 1 minute intervals will look fine when played back at 30 fps.

The rate at which its shot plays a role in how much you have to slow it down in post-processing, I believe.

[jagabo]

I installed the plugin and .MKV does show up.

If you have multiple filters that tell VirtualDub they can open MKV files the program may not be using the filter you think it is. Use the Files Of Type pulldown to specify which source filter (MKV? DirectShow? ffmpeg?) should be used. To use the DirectShow source filter you must have DirectShow components installed that will allow the reading and splitting of MKV files, and decompression of whatever codecs you used. I think the MKV source filter will require that you have VFW codecs for the codecs you used (not sure about this though). The ffmpeg source filter is self contained but I think it's buggy.

Unless there is another program that can take ACCURATE screen captures, unlike Media Player Classic (anyone?)

MPC probably isn't the problem. Your graphics card is probably performing some post processing that's screwing up the image. Are you using the Print Screen key or MPC's File -> Save Image? They work differently.

I don't really see the difference between JPEG and RAW in terms of quality.

If you zoom into JPEG images you'll see blocky artifacts that aren't in the raw images. That's probably not an issue for what your doing.

If I render it at 15fps and speed it up it won't have the same effect as if I render it at 60fps but slow it down.

Unless you're performing some kind of frame blending or motion interpolation the result will be the same. If you're using bitrate based encoding the faster frame rate will deliver lower quality.

[theewizard]

i would resize the pictures using photo software
before working on the time laspe video

save the resized pix at 100% quality perhaps in png, or other format that has less compression loss than jpeg
maybe even 'raw'

[Cornucopia]

OK, maybe you are new to working with video, so I'll try to bring you up to speed (pardon the unintended pun):

Unlike photos, video doesn't easily upgrade to the next resolution step easily. With photos, you can easily go from a 4MegaPixel cam to 16 to 64 if you want to make use of the increased resolution (though you should already know that MP really isn't the only deciding or maybe even MAIN factor in providing quality - there's things such as lens resolving power, sensor size and sensitivity, dynamic range, etc).
With video, the world used SD resolution (~720x480 or ~720x576)for decades until the late 1990s when things started moving to HD (1280x720 or 1920x1080). This HAS accellerated somewhat since, including adding 2k & 3k & 4k digital cinema resolutions, and quite very recently the addition of common UHD 4k & 8k and digital cinema 5k resolutions, with 12k coming down the way. But those are few and far between compared to the expansion of photo resolution (which already had started out ahead of video).
Video has other things going for it that photos don't have, namely MOTION.

However, it's quite unrealistic to keep something as a "5k" video unless you have something to play it on (smoothly) and something to display it on (clearly & accurately). Even your "test" timelapse, you aren't actually seeing correctly or natively, because you PC is (behind the scenes) doing a continuous resize down to the resolution of your screen (which is only ~"4k"). Note that the resizing is also inherently losing quality just from the interpolation alone, even with Uncompressed/Lossless sources & destinations.

If you intend to distribute this, you're going to need to pick a COMMON consumer resolution. If you can afford to project this at a cinema and you want to maintain your highest resolution, you could downsize to 4k or 2k and create a DCI-compliant DCP disc(s) set. Next down from that would be to go 1920x1080 on Blu-ray. Or, if this is intended purely for some scientific presentation or experimental excersize, you could upload it and have those scientists download & evaluate the media in it's native resolution, but even then, like you, they would have to either downsize to a similar resolution as your own (which happens to be the LARGEST common computer display resolution) as full-screen, or keep it 100% size and only view a cropped portion in a window.

Now, back to motion...

The "smoothness" and "fluidity" of a motion is proportional to its unique frame's framerate. This is not to say that the motion must be faster. An analog clock's second hand moving around in a circle will look MUCH more fluid at 30fps vs. 15fps, and much more fluid still at 60fps. Even though all of these versions would be running for the same duration. The fluidity is due to the smaller stepsize in the change of placement within the objects' motion.

Timelapse is creating motion out of stills. It is imposed in the way you desire to evoke the scene. A flower petal opening can look very different depending upon both how often it is captured and what the final framerate is. There is of course a ratio involved here.
A TL video running at 30fps of something recorded at 1fps will run 30x faster. It will be more fluid by 5 levels of magnitude (doubling each time) but will be 30x shorter in duration. So if you shot your 1fps stills for 30 minutes, your final video will last only 1 minute.

Only you can determine your fluidity & duration, but that's what you need to figure out FIRST, not after the fact, because that determines the tenor of the presentation.

Importantly: the determination of codec should NOT be your first priority. It follows from your general conversion/processing workflow, which follows from your project design.

****************************

Your steps heretofore have been kind of haphazard from a video tech's standpoint. I would say it would be a very good idea to take a step back and decide how you want your project to look and to whom you are going to show it to. Using that, and letting us in on your decisions, you can then map out what would be the best course of action to retain the best possible (available) quality. Notice you may have to give up certain assumptions that are more photo-centric and not so conducive to good, workable video.

I can tell you that there are much better processes to use than "VideoMach". When I think of the better editors/processors that can work with still image sequences (and create timelapse), it doesn't even get on the radar!

Scott

[Track]

If you have multiple filters that tell VirtualDub they can open MKV files the program may not be using the filter you think it is. Use the Files Of Type pulldown to specify which source filter (MKV? DirectShow? ffmpeg?) should be used. To use the DirectShow source filter you must have DirectShow components installed that will allow the reading and splitting of MKV files, and decompression of whatever codecs you used. I think the MKV source filter will require that you have VFW codecs for the codecs you used (not sure about this though). The ffmpeg source filter is self contained but I think it's buggy.

You were right! I got it to work by placing the "plugins32" folder in the root (not in the plugins folder) and deleting all but the DirectShow Input Driver. The 64-bit one did not work.

MPC probably isn't the problem. Your graphics card is probably performing some post processing that's screwing up the image. Are you using the Print Screen key or MPC's File -> Save Image? They work differently.

I'm using MPC's "Save Image."

If you zoom into JPEG images you'll see blocky artifacts that aren't in the raw images. That's probably not an issue for what your doing.

Here is the difference (at 500%):

This is the JPEG captured with MPC. It looks pixelated for some reason (doesn't look that way when viewing it), but there are obvious artefeacts around the building.


This is the original file.


This is the same file as the first, but captured with VirtualDub instead of MPC.

Unless you're performing some kind of frame blending or motion interpolation the result will be the same. If you're using bitrate based encoding the faster frame rate will deliver lower quality.

You're saying that if I render at 15fps and speed it up it will look the same as if I render at 30fps and speed it down?

So, what framerate should I choose if I can only choose one?

Your steps heretofore have been kind of haphazard from a video tech's standpoint. I would say it would be a very good idea to take a step back and decide how you want your project to look and to whom you are going to show it to. Using that, and letting us in on your decisions, you can then map out what would be the best course of action to retain the best possible (available) quality. Notice you may have to give up certain assumptions that are more photo-centric and not so conducive to good, workable video.

Well, like I said, I want one "negative" uncompressed video, so that I can take out individual frames at the same quality as the original files, and from that I'll likely render at 1080p or 4k. It doesn't really matter what I do if I intend to upload it to Youtube, because the quality will always be awful.

I can tell you that there are much better processes to use than "VideoMach". When I think of the better editors/processors that can work with still image sequences (and create timelapse), it doesn't even get on the radar!

Please, I would love to know! It seems as if I am stuck using XviD4PSP, but I can use any timelapse software I wish.

Although, I'm quite fond of both VideoMach and VirtualDub.

[poisondeathray]

This is the JPEG captured with MPC. It looks pixelated for some reason (doesn't look that way when viewing it), but there are obvious artefeacts around the building.

Because it's JPEG. JPEG is lossy compression. Use BMP or PNG instead to take the screenshot

In the save image dialog box, "save as type" there is a drop down box

[Track]

This is the JPEG captured with MPC. It looks pixelated for some reason (doesn't look that way when viewing it), but there are obvious artefeacts around the building.

Because it's JPEG. JPEG is lossy compression. Use BMP or PNG instead to take the screenshot

In the save image dialog box, "save as type" there is a drop down box

But the other two images are also JPEG..

[poisondeathray]

This is the JPEG captured with MPC. It looks pixelated for some reason (doesn't look that way when viewing it), but there are obvious artefeacts around the building.

Because it's JPEG. JPEG is lossy compression. Use BMP or PNG instead to take the screenshot

In the save image dialog box, "save as type" there is a drop down box

But the other two images are also JPEG..

JPEG has variable quality . Likely you are using different quality levels when taking the screenshot. Many programs have a slider for the quality 0-100. AFAIK, the quality level isn't adjustable in MPC/HC . To rule out other variables confounding the conclusions you draw from observations, you always use lossless comparisons when possible

If your source files are RGB (no colorspace conversions), and you took lossless screenshots, but there still was a difference - then you might look at considering other issues

[Track]

This is the JPEG captured with MPC. It looks pixelated for some reason (doesn't look that way when viewing it), but there are obvious artefeacts around the building.

Because it's JPEG. JPEG is lossy compression. Use BMP or PNG instead to take the screenshot

In the save image dialog box, "save as type" there is a drop down box

But the other two images are also JPEG..

JPEG has variable quality . Likely you are using different quality levels when taking the screenshot. Many programs have a slider for the quality 0-100. AFAIK, the quality level isn't adjustable in MPC/HC . To rule out other variables confounding the conclusions you draw from observations, you always use lossless comparisons when possible

If your source files are RGB (no colorspace conversions), and you took lossless screenshots, but there still was a difference - then you might look at considering other issues

No, I saved all of them with MS Paint. No difference at all.

[poisondeathray]

No, I saved all of them with MS Paint. No difference at all.

Did save them with MPC then brought them into MS Paint ? Clarify what you're doing exactly

I'm using MPC's "Save Image."

When you save with MPC use lossless PNG or BMP , then bring them into MS Paint

[jagabo]

This is the JPEG captured with MPC. It looks pixelated for some reason (doesn't look that way when viewing it), but there are obvious artefeacts around the building.

Those are a result of a sharpening filter. Probably by the graphics card. Go to it's setup applet and disable all the "enhancement" features. All they do is screw up video. MPC also has such filters. Disable those too.

Unless you're performing some kind of frame blending or motion interpolation the result will be the same. If you're using bitrate based encoding the faster frame rate will deliver lower quality.

You're saying that if I render at 15fps and speed it up it will look the same as if I render at 30fps and speed it down?

Let's say you have 600 frames. If you make a video with all 600 frames at 60 fps you will have a video that plays for 10 seconds. If you make another video with all 600 frames at 15 fps you will have a video that plays for 40 seconds. If you play the first video at 30 fps you will still see all 600 frames but they will play over a period of 20 seconds. If you play the second video at 30 fps you will also see all 600 frames and the video will also take 20 seconds. The two will look exactly the same (aside from bitrate considerations).

Whether any software you use to perform similar frame rate conversions does the same will depend on the software and the settings you use.

So, what framerate should I choose if I can only choose one?

You choose a bitrate that gives the "look" you want, and that match any requirements for your intended output device(s). Blu-ray is limited to 60 fps max. Your computer doesn't have a specific limit but it won't be able to smoothly play back large frame sizes at 1000 fps. And even if it could you would only see 60 different frames per second on your computer monitor.

I can tell you that there are much better processes to use than "VideoMach".

Please, I would love to know! It seems as if I am stuck using XviD4PSP, but I can use any timelapse software I wish.

Although, I'm quite fond of both VideoMach and VirtualDub.

VirtualDub can decimate or duplicate frames for frame rate conversions. Just use the lower part of the frame rate dialog. The biggest limitation with VirtualDub is the fact that it really only outputs AVI files. (It does support external encoders and muxers though.)

There are more advanced ways of performing frame rate conversions. Some use motion interpolation to generate in-between frames. Ie, they will see that an object moved from point A to point B between two frames and double the frame rate by creating a new frame with that object halfway between A and B. This works well for some types of material, not so well for others. Some programs blend frames to change the frame rate. To cut the frame rate in half the will blend pairs of frames together. To double the frame rate they'll create in-between frames that are a blend of the frame before and after. This may look a little smoother than simply throwing out every other frame, at the cost of blurry looking frames when there's motion.

[Track]

When you save with MPC use lossless PNG or BMP , then bring them into MS Paint

Holy shit, that worked!

Saving through MPC as PNG, then converting it to JPEG looks identical to the original file.

But how is that possible..?

Taking 2GB of photos, processing them into a 115MB video, then capturing a single frame as PNG looks EXACTLY THE SAME as the actual photos from the 2GB folder!

Can someone explain this?

Let's say you have 600 frames. If you make a video with all 600 frames at 60 fps you will have a video that plays for 10 seconds. If you make another video with all 600 frames at 15 fps you will have a video that plays for 40 seconds. If you play the first video at 30 fps you will still see all 600 frames but they will play over a period of 20 seconds. If you play the second video at 30 fps you will also see all 600 frames and the video will also take 20 seconds. The two will look exactly the same (aside from bitrate considerations).

Whether any software you use to perform similar frame rate conversions does the same will depend on the software and the settings you use.

You choose a bitrate that gives the "look" you want, and that match any requirements for your intended output device(s). Blu-ray is limited to 60 fps max. Your computer doesn't have a specific limit but it won't be able to smoothly play back large frame sizes at 1000 fps. And even if it could you would only see 60 different frames per second on your computer monitor.

Wow, I didn't realize that, but it seems perfectly logical.

So, you're saying that it doesn't matter what frame-rate I choose?

But considering that I am going to delete the photos, I need to pick a frame-rate that would work as a "negative", meaning I could make it slower or faster and it would be the same if I rendered it at 15fps or 60fps.

[jagabo]

But how is that possible..?

Taking 2GB of photos, processing them into a 115MB video, then capturing a single frame as PNG looks EXACTLY THE SAME as the actual photos from the 2GB folder!

It's not exactly the same if you used a lossy codec. But something to consider: much of the compression of video comes from not repeating parts of the picture that don't change from frame to frame. So in a talking head shot with a static background, the first frame might be encoded much like a JPEG picture (the entire picture is encoded) but many subsequent frames will only encode the moving lips/head of the speaker.

But considering that I am going to delete the photos, I need to pick a frame-rate that would work as a "negative", meaning I could make it slower or faster and it would be the same if I rendered it at 15fps or 60fps.

It doesn't matter what frame rate you choose as long as the video contains all the frames. The frame rate is just a number in the header that tells the player (or editor) how long each frame should be displayed during playback.

[Track]

It's not exactly the same if you used a lossy codec. But something to consider: much of the compression of video comes from not repeating parts of the picture that don't change from frame to frame. So in a talking head shot with a static background, the first frame might be encoded much like a JPEG picture (the entire picture is encoded) but many subsequent frames will only encode the moving lips/head of the speaker

But the question is - can I delete the original files and re-capture them again from the 100MB video at the same quality, if I ever want to?

If not, can I do this with the 2GB video?

It doesn't matter what frame rate you choose as long as the video contains all the frames. The frame rate is just a number in the header that tells the player (or editor) how long each frame should be displayed during playback.

And this is changeable?

[jagabo]

But the question is - can I delete the original files and re-capture them again from the 100MB video at the same quality, if I ever want to?

Not if you used a lossy codec.

If not, can I do this with the 2GB video?

In which you used a lossless codec? Yes.

It doesn't matter what frame rate you choose as long as the video contains all the frames. The frame rate is just a number in the header that tells the player (or editor) how long each frame should be displayed during playback.

And this is changeable?

Yes. With AVI files you can use AviFRate. Or remux with VirtualDub after changing the source frame rate in the frame rate dialog. You can even use a hex editor of you know what you're doing. With other containers (MKV, MP4) you can usually set the frame rate within remuxing tools. The MPG container only supports a few fixed frame rates but you can easily change the flagged frame rate within those limitations.

In addition, most editors will let you override the frame rate after importing the video. Ie, you can tell the editor to ignore the frame rate flagged in the source and assume another frame rate. That's what you're doing in the top part of VirtualDub's frame rate dialog. In an AviSynth script you can simply specify AssumeFPS(x) where x is the frame rate you want.

[Track]

Thank you everyone, and jagabo in particular. I'll write here again if I have another question, but thank you!

EDIT: I'm having a serious problem with VirtualDub. I'm trying to render another timelapse and it just hangs after the 20-something frame. The ETA clock keeps going, estimating the time to be more and more, and nothing happens. When I try to cancel it, it hangs for a little bit.

[_Al_]

Working with video you have to know what it is made for and you choose workflow. Because of those things like frame rate, resolution, choosing complexity of compression, and device a person will watch the video with.

Maybe would be better if you say what do you have your video for. Best home video? Best Youtube upload? Best for web streaming etc.

[jagabo]

Is there a program that will mux a series of JPEG images into an MJPEG video? That would seem like an optimal solution. It sounds like mencoder might be able to do this:

http://www.mplayerhq.hu/DOCS/HTML/en/menc-feat-enc-images.html

Creating a Motion JPEG (MJPEG) file from all the JPEG files in the current directory:
mencoder mf://*.jpg -mf w=800:h=600:fps=25:type=jpg -ovc copy -oac copy -o output.avi

Yes, I verified that it works. The AVI was approximately the same size as the sum of the JPG files and the mux was very quick. Obviously, you have to set the width and height parameters to the correct values for your JPEG files, and the fps to whatever you want.

[poisondeathray]

ffmpeg is another option that can use an image sequence for input , and output to a variety of formats

If this was for computer only , I would leave them as an image sequence, play them back with an avs script in avisynth ImageSource() . The benefit is original quality, same size as original instead of a huge 10x lagarith file. That original sequence is your "negative" , and you can encode to a variety of formats, resize... whatever... with avisynth + some encoder

[Track]

I tried both Mencoder and ffmpeg. They both exit upon start and I can't seem to read the message as to why..

[jagabo]

Start them from a Command Prompt window. Start -> All Programs -> Accessories -> Command Prompt. That way the window won't disappear after running the program.

Or run it from a batch file with a "pause" afterward:

mencoder mf://*.jpg -mf w=800:h=600:fps=25:type=jpg -ovc copy -oac copy -o output.avi
pause