Rendering Uncompressed - Still looks worse than original..

[Track]

I'm rendering a timelapse video out of a number of still shots.

I'm using VideoMach because it is the best software I've found so far (if you know something better, I'm all ears).

I tried rendering Full-Frame uncompressed, but the file limit is 2GB for some reason.

So, I'm using Xvid. I've set everything to unrestricted and have tweaked the settings (as best as I know how) a thousand times.

However, there is still a degradation in quality.

Here is an example:

This is the original.



This is an image taken from the video encoded.



This is at 250%. I know that at 100% you can't tell the difference, but I'm picky and I don't want to delete the original files without having something that could re-create the same quality if need be.

If anyone can tell me what I'm doing wrong, I'd very much appreciate it.

Thank you.

[DB83]

2 gig file limit !!!. What OS are you using ?

However you tweak xVID the video will still be compressed. You could try a lossless codec such as Lagarith but a 2 gig file limit !!!!!!!!

[Track]

So, how can you render uncompressed without a 2GB file limit?

Doesn't the file limit make rendering uncompressed futile?

[Selur]

However, there is still a degradation in quality.

That's normal since:
a. Xvid is a MPEG-4 ASP based encoder and MPEG-4 ASP is never lossless.
b. you do a color space conversion from RGB (4:4:4) to Yv12 (4:2:0), see: https://en.wikipedia.org/wiki/YUV

So, how can you render uncompressed without a 2GB file limit?

the point is normally not to render uncompressed but lossless
-> there are multiple lossless video codecs which are ment as intermediate formats.

----
2GB is the limit either because it's an AVI created through vfw (see: http://neuron2.net/LVG/filesize.html, most applications nowadays support opendml avi files which do not have this restriction), or due to a design flaw of the tool used.

[Track]

b. you do a color space conversion from RGB (4:4:4) to Yv12 (4:2:0), see: https://en.wikipedia.org/wiki/YUV

the point is normally not to render uncompressed but lossless
-> there are multiple lossless video codecs which are ment as intermediate formats.

So, if I don't convert the color space, it will look the same?

If not, where do I find lossless video codecs? All I know is where to find decoding codec (K-Lite Codec Pack).

Actually, I've always wanted to know more about encoding codec. I find myself always feeling like I'm converting or re-encoding certain video files and losing quality, no matter what I do.

I'd love to know that I was encoding MKV at the highest quality-to-filesize ratio.

Thanks again.

[jagabo]

So, if I don't convert the color space, it will look the same?

Converting Xvid to Xvid will always result in a video that looks worse than the original.

Take a smooth piece of aluminum foil. Crush it into a ball (compress it). Now open it up (decompress it). Notice all the wrinkles? Now crush it into a ball again. And open it up again. Did it get more wrinkles?

[DB83]

You were shown about the lossless codec in my reply above - just click the link. Here it is again >>>>> Lagarith.

But it seems that the program is flawed - I just took a quick look at it - there is a 'no-splitting' option but I doubt if that applies to avi. It is not BTW a 2 gig limit but the file will be split into 2 gig sections but of no use unless you can join them up.

And, to be honest, any program that suggests you download a codec pack really should be avoided.

[Selur]

So, if I don't convert the color space, it will look the same?

if you use a lossless codec: yes (Mpeg-4 ASP only supports Yv12 and is by design never lossless)

If not, where do I find lossless video codecs?

about lossless codecs, see: http://www.compression.ru/video/codec_comparison/lossless_codecs_2007_en.html
most common lossless codecs from my point of view nowadays are:

• Ut Video Codec Suite
• Canopus LossLess Codec
• MSU Lossless Video Codec
• Lagarith
• FFV1
• HuffYuv
• x2645 lossless
• camstudio lossless
• ... there are tons of others

normally lossless codec only differ by:
a. encoding speed
b. decoding speed
c. compression ratio
-> if one is good at least one other normally suffers + not all tools support all codecs

I'd love to know that I was encoding MKV at the highest quality-to-filesize ratio.

if it's not for pc only playback using a lossless codec is probably the last you want to do,...
Also lossless codecs are normally ment as intermediate and are huge in file size.

-> if you state more detailed what you are trying to do, some of us here might post some suggestions regarding codec choice.

Cu Selur

[jagabo]

Recompressing a video involves first decompressing it to uncomrpessed frames, then compressing with the new codec. Converting from Xvid to a lossless codec like Lagarith will usually cause the file to grow in size by 5 to 50 fold.

[Track]

Recompressing a video involves first decompressing it to uncomrpessed frames, then compressing with the new codec. Converting from Xvid to a lossless codec like Lagarith will usually cause the file to grow in size by 5 to 50 fold.

I think there has been a misrepresentation on my part.

I am not converting Xvid to Xvid.

I am creating a Time-Lapse video from photo stills using the program VideoMach.

Would it be possible to load the lossless codec into it?

EDIT: I have successful implemented Lagarith into VideoMach.

However, the quality is still exactly the same. There is a noticeable difference at 250%.

[jagabo]

I am not converting Xvid to Xvid.

I am creating a Time-Lapse video from photo stills using the program VideoMach.

Would it be possible to load the lossless codec into it?

Yes. I don't know anything about VideoMach but if your photos are sequentially numbered I would use VirtualDub for this. It's free and easy to use. Just open the first photo (File -> Open Video File) and it will automatically append all the sequentially numbered photos. Set the desired frame rate (Video -> Frame Rate). Then set select the codec and configure it (if necessary, Video -> Compression, Configure). And finally save the file (File -> Save as AVI).

[Track]

I am not converting Xvid to Xvid.

I am creating a Time-Lapse video from photo stills using the program VideoMach.

Would it be possible to load the lossless codec into it?

Yes. I don't know anything about VideoMach but if your photos are sequentially numbered I would use VirtualDub for this. It's free and easy to use. Just open the first photo (File -> Open Video File) and it will automatically append all the sequentially numbered photos. Set the desired frame rate (Video -> Frame Rate). Then set select the codec and configure it (if necessary, Video -> Compression, Configure). And finally save the file (File -> Save as AVI).

Okay, I followed your instructions and used Lagarith. The file size was a little over 2GB, which is LARGER than all the photos combined in a folder, and it STILL has artefacts/blockiness.

It's odd.. the overall file size is 2GB but when I capture a single image frame, it's 1/5 the size of the actual photo.

This is starting to become quite perplexing.

EDIT: I honestly don't know what I'm doing wrong..

I decided to render everything completely uncompressed, and the file size was 16GB(!)

However, upon capturing a single frame, it's no different than the 2GB file.

[jagabo]

Okay, I followed your instructions and used Lagarith. The file size was a little over 2GB, which is LARGER than all the photos combined in a folder

Jpeg photos? Remember they are compressed with a lossy codec. Decompressing them and saving with a lossless codec is expected make them grow, usually.

and it STILL has artefacts/blockiness.

That wasn't caused by the Lagarith compression. Look elsewhere.

but when I capture a single image frame, it's 1/5 the size of the actual photo.

How big the losslessly compressed frame is varies depending on what's in the photo. If there's little detail and little noise you might get very small file. If there's lots of detail and noise you will get a large file.

[Track]

When I encode the JPEGs with Lagarith, the outcome simply looks worse.

What can I do to keep the same quality?

Essentially, all I want is to combine all the photos into one video file that is as large as the sum size of all the photos that make it up.

200MB worth of photos should transfer to a 15fps video file that is also around 200MB. That way, I render my timelapse and have the ability to capture individual photos from the frames.

[jagabo]

When I encode the JPEGs with Lagarith, the outcome simply looks worse.

Again, that's not caused by Lagarith. The problem is somewhere else in your processing.

[jagabo]

Essentially, all I want is to combine all the photos into one video file that is as large as the sum size of all the photos that make it up.

You probably won't achieve that with Lagarith. Use a lossy codec like MJPEG or Xvid. But you will cause some loss of quality.

[Track]

When I encode the JPEGs with Lagarith, the outcome simply looks worse.

Again, that's not caused by Lagarith. The problem is somewhere else in your processing.

Well, what else could it be?

The photos are stored on my SSD.
I've made the changes in VirtualDub that you've instructed.
I save to .AVI using Lagarith.

I've tried different settings within Lagarith, but the outcome is always the same.

What I'm saying is, I'm not doing anything else to process the images, so I don't know what else it could be..

Perhaps the frame capture in Media Player Classic is flawed..?

I'm honestly stumped.

[jagabo]

What kind of problems are you seeing now? MPC could indeed be the problem. Open your Lagarith AVI with VirtualDub. How does that look compared to opening your JPEG files in VirtualDub? (You can run mulitple instances of VirtualDub at the same time.)

In VirtualDub try forcing the input and output colorspaces to YV12 (JPEG photos normally use 4:2:0 chroma subsampling) using Video -> Color Depth. Configure Lagarith to work in YV12. See if that makes any difference.

[vhelp]

may i ask.. what camera make and model series are you using ?

bc u might have a "raw" format.. that is lossless. then take that and transfer it to a lossless codec. i believe the UT code suite (was recently updated) can give you 4:4:4 output. you would need to install that codec and then your sw will (should) open that.

edit: oh, just realized.. these are pics already compressed to jpeg by your camera. but, moving forward, you should set your camera to "raw" if it has that feature.

[jagabo]

PicVideo's MJPEG codec has a 4:4:4 mode (you have to feed it RGB and the output will be RGB). If I remember correctly they call it "1:1:1". But even cameras with "raw" modes don't really produce 4:4:4 chroma subsamples. The bayer pattern of the sub pixels in the sensor results in lower color resolution than luma resolution. 4:4:4 encoding of the raw source might still give slightly better results though.

[Track]

What kind of problems are you seeing now? MPC could indeed be the problem. Open your Lagarith AVI with VirtualDub. How does that look compared to opening your JPEG files in VirtualDub? (You can run mulitple instances of VirtualDub at the same time.)

I can't view the file in VirtualDub because it's an 18MP file and I zoom in 650%. I haven't found a way to pan the video, so it only shows me the top left corner. There is also no way of capturing and saving a single frame (that I could see.)

In VirtualDub try forcing the input and output colorspaces to YV12 (JPEG photos normally use 4:2:0 chroma subsampling) using Video -> Color Depth. Configure Lagarith to work in YV12. See if that makes any difference.

There is no difference whatsoever.

[Track]

may i ask.. what camera make and model series are you using ?

bc u might have a "raw" format.. that is lossless. then take that and transfer it to a lossless codec. i believe the UT code suite (was recently updated) can give you 4:4:4 output. you would need to install that codec and then your sw will (should) open that.

edit: oh, just realized.. these are pics already compressed to jpeg by your camera. but, moving forward, you should set your camera to "raw" if it has that feature.

I would, but timelapses take thousands of photos and my memory card can only take so much.

[jagabo]

I can't view the file in VirtualDub because it's an 18MP file and I zoom in 650%. I haven't found a way to pan the video, so it only shows me the top left corner. There is also no way of capturing and saving a single frame (that I could see.)

Video -> Copy Source Frame To Clipboard. Then in a photo editor paste as a new image.

Or if you use VirtualDubMod (it's outdated but useful for this), select Video -> Snapshot Source Frame then save as BMP or PNG.

[DB83]

[QUOTE=Track;2237271]When I encode the JPEGs with Lagarith, the outcome simply looks worse.
.
.


200MB worth of photos should transfer to a 15fps video file that is also around 200MB. QUOTE]

Video does not work like that.

The filesize depends on the run-length of the video and the bit-rate.

so you could have 400 mb of photos but that does not mean that the video will be twice the size.

On a more technical point, I do not think I read how many photos you have that make up the 200 meg but they inevidably will be low quality. How many mega-pixels ?. Applying video codecs to these might well amplify that matter. You really must start off with high quality photos.

[Track]

On a more technical point, I do not think I read how many photos you have that make up the 200 meg but they inevidably will be low quality. How many mega-pixels ?. Applying video codecs to these might well amplify that matter. You really must start off with high quality photos.

I'm not sure what you mean "High Quality Photos". I shoot with my T2i and a 500$ lens.

Video -> Copy Source Frame To Clipboard. Then in a photo editor paste as a new image.

Or if you use VirtualDubMod (it's outdated but useful for this), select Video -> Snapshot Source Frame then save as BMP or PNG.

Wow, I don't believe it..

Using VirtualDub, the capture looks identical to the photo!

Wow, I can't believe Media Player Classic would do this

Btw, I saved as JPEG, not BMP or PNG. Is there a difference?

[jagabo]

Btw, I saved as JPEG, not BMP or PNG. Is there a difference?

JPEG is a lossy compression technique that usually works with 4:2:0 chroma subsampling. BMP is usually uncompressed RGB (and hence lossless). PNG is usually losslessly compressed RGB.

[DB83]

Well I do not have a $500 lens on my DSLR. But I do know what mega-pixels my camera is capable of. Do you ?

I was just asking how many pics made up that 200 meg? I can get over 900 on a 4 gig card . Lower the quality (nothing to do with the lens) and I can get more.

But let us get back to video. I see you create a 15 fps video. Is that wise ?

[vhelp]

The bayer pattern of the sub pixels in the sensor results in lower color resolution than luma resolution.

i forgot to mention that. and, there is an avisynth plugin for this if i recall. not sure it would help the op.

@ op, then save as png.

[Track]

Well I do not have a $500 lens on my DSLR. But I do know what mega-pixels my camera is capable of. Do you ?

Well, it wouldn't make much sense to own a 500$ lens if I didn't know that.

As I said above, it's a Canon T2i, so it shoots 18MP.

I was just asking how many pics made up that 200 meg? I can get over 900 on a 4 gig card . Lower the quality (nothing to do with the lens) and I can get more.

Lower the quality? It's already low enough as it is as a JPEG.

But let us get back to video. I see you create a 15 fps video. Is that wise ?

I honestly have no idea. My next step is to figure it out.

Should I render 30fps and then slow it down with Sony Vegas? That would be more smooth, no?


I am also wondering two things:

1. If I'm going to slow the video down, I'll have to essentially re-encode it, which would lower the quality, no?

2. Is there another piece of software, like VirtualDub, that takes ACCURATE screen grabs? VirtualDub for some reason does not support the files output from XviD4PSP..

[DB83]

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.