Captured AVI file too large

[jagabo]

The work you 've done here is EXCELLENT!!! Show me the steps please!

It did need deinterlacing after all eh?

No, it didn't need deinterlacing but you've seemed to want it so I used the best deinterlacer I know of. Attached is an interlaced MPEG 2 video (no audio) with minimal processing (just fixing the striped colors and cleaning the borders). I used the following AviSynth script:

AVISource("Test_1Min.avi")
AssumeTFF()
ConvertToYV12(interlaced=false)
Crop(8,0,-8,-16)
AddBorders(0,8,0,8)
BilinearResize(352,480)

To fix the striped colors problem I use ConvertToYV12(interlaced=false) (interlaced=true would normally be used for interlaced video). I cropped away the small black borders at the left and right edges, and the noise at the bottom of the frame. Then added black borders at the top and bottom of the frame to bring height back up to 480 lines. I resized to 352x480 in preparation for half-D1 MPEG 2 encoding. I used HcEnc via HcGUI to encode as MPEG 2 with an average bitrate of 3000 kbps, max 4500 kbps. That will allow you to put about 6 hours on a dual layer DVD. Encoding took about 1 minute with HcEnc via HcGUI.

I too have a a 2.4 GHz dual core, INTEL though, 2 gb RAM...so how much processing time for a 6 hour tape?

About 240 hours if you include both TempGaussMC_beta1mod() and Neat Video. About 6 hours if you did the minimal filtering of my second example.

[drgt]

No, it didn't need deinterlacing but you've seemed to want it...

I did not actually WANT it. I was merely quoting those other sources which recommended deinterlacing.

1. How can you tell if the file to be processed is interlaced or not? Then, if it is, how can you tell it needs deinterlacing? And then how you determine the method to be used?

2. Would the quality be just as good if you did not do that step?


Certainly, the time saved in the second encoding is worth considering. I never used Avisynth or HcGui; I hope I will be able to deploy your instructions.

(interlaced=true would normally be used for interlaced video).

The source or the target? If "source" is the answer, I thought you said the source is interlaced. If "target" is the answer, you are applying deinterlacing which, you said, is not needed. I am still confused about this issue...

[jagabo]

I did not actually WANT it. I was merely quoting those other sources which recommended deinterlacing.

About the only times I would recommend deinterlacing video is if you are going to send it to Youtube (or some other online video sharing site) or if you are making AVI files to send to people who may not know how to enable deinterlacing in their player.

1. How can you tell if the file to be processed is interlaced or not? Then, if it is, how can you tell it needs deinterlacing? And then how you determine the method to be used?

https://forum.videohelp.com/threads/325830-Captured-AVI-file-too-large?p=2020098&viewfu...=1#post2020098

If every pair of frames is (nearly) the same you have progressive video (ie, not interlaced)

There is an exception to this (normally found in PAL video) but I don't want to get into it here.

2. Would the quality be just as good if you did not do that step?

Deinterlacing always introduces artifacts. Whether you do it before encoding or rely on the player to deinterlace. The difference is, if you deinterlace before encoding the artifacts become a permanent part of your video. If you encode interlaced and let the player deinterlace, it may not look as good as TempGaussMC_beta1mod() now, but in the future better deinterlacers may arise so your playback of your interlaced video will improve.

Certainly, the time saved in the second encoding is worth considering.

TempGaussMC_beta1mod() has some faster (about 5x) settings. But they don't deliver quite as good quality. And "faster" than glacially slow is still pretty slow!

(interlaced=true would normally be used for interlaced video).

The source or the target? If "source" is the answer, I thought you said the source is interlaced. If "target" is the answer, you are applying deinterlacing which, you said, is not needed. I am still confused about this issue...

As always, we're talking about the state of the video at the time the filter is applied. In this particular case we have an interlaced source -- but the chroma channels are encoded progressive. This is a bit hard to explain, just take my word for it that ConvertToYV12(interlaced=false) fixes the chroma channels.

[drgt]

So, the only reason in my case you deinterlaced is that the chroma channels were messed up. Otherwise, you would just first resize and crop and perhaps apply the neat video filter second right?

As far as interlaced status identification, per your advise, I did advance through the frames using the "Bob Doubler - TFF" filter and, like I said, to me every frame was different from the previous, not 2 frames were near the same. With BFF though, it gave me a back and forth movement.

Could you please post the neat video configuration settings and direct me on using the Avisynth with the script you used?

Thanks

How do I start this thing?

[jagabo]

So, the only reason in my case you deinterlaced is that the chroma channels were messed up.

Partly that, and also because it was easier to filter and encode as Xvid. And TempGaussMC's double frame rate output would show the same smoothness of motion and sharpness of the individual frames (compared to Field Bob - Smooth's blurry frames and jerkier motion).

Otherwise, you would just first resize and crop and perhaps apply the neat video filter second right?

You have to be careful with interlaced video. You need to make sure the two half-images encoded into each frame don't co-mingle. And you have to keep track of the field order. Even numbered scanlines need to remain even, odd numbered scanlines need to remain odd. If not, you have reversed the field order and must encode accordingly.

As far as interlaced status identification, per your advise, I did advance through the frames using the "Bob Doubler - TFF" filter and, like I said, to me every frame was different from the previous, not 2 frames were near the same. With BFF though, it gave me a back and forth movement.

Yes, your video is interlaced and TFF. When watched on TV you need to see the top field first, then the bottom field. Otherwise you will get that two-steps-forward-one-step-back jerky motion.

Could you please post the neat video configuration settings

I don't remember exactly what settings I used and it's not really possible to post them exactly anyway. Part of the process is to give Neat Video a sample that's supposed to be a flat area so it can identify the nature of the noise. Locate a frame with a large flat area and use the Auto Profile option. Neat video should draw a blue box where it finds a suitable part of the frame to sample:



If you don't like you area it picks, mark a box yourself. For example, in that sample image the bottom field has much more noise than the top field*. It might be better to sample that field instead. You can drag the blue box down to the bottom field or mark an area yourself.

If the video is interlaced be sure the interlaced option is selected on the main Neat Video dialog. The Temporal Filter setting helps get rid of noise from frame to frame. If you set it too high small details will get lost when there is motion -- for example a fuzzy wool sweater may look like smooth silk.

and direct me on using the Avisynth with the script you used?

Basically, you use a text editor (like Notepad) to create the AviSynth script. I usually put the script in the same folder as the video file I'm working on. When you save the script be sure to give it the extension .AVS, not .TXT. Then in VirtualDub open the .AVS file, not the video file.

More detailed instructions can be found in the AviSynth docs.
http://avisynth.org/mediawiki/Main_Page#New_to_AviSynth_-_start_here

* I noticed that the top field of your video is cleaner and darker than the bottom field. For best results you would want to filter the two fields separately then weave them back together. You can't do that with VirtualDub though. And it's pretty advanced.

[drgt]

You have to be careful with interlaced video. You need to make sure the two half-images encoded into each frame don't co-mingle. And you have to keep track of the field order. Even numbered scanlines need to remain even, odd numbered scanlines need to remain odd. If not, you have reversed the field order and must encode accordingly.

I do all that in Virtualdub? Where? How?

As far as interlaced status identification, per your advise, I did advance through the frames using the "Bob Doubler - TFF" filter and, like I said, to me every frame was different from the previous, not 2 frames were near the same. With BFF though, it gave me a back and forth movement.

Because not 2 frames are alike, I would have thought I have a progressive video.

If I have several capture pieces from the same tape is it best to append them in VirtualDub before processing or after? If before, should I make a new "raw" file containing all pieces and then process this file?

****With all the knowledge you have and input from this site you could put together a video editing manual!!! Someone ought to write a manual for VirtualDub, but as the author says, it would take perhaps more time than writing the program code.****

[jagabo]

You have to be careful with interlaced video. You need to make sure the two half-images encoded into each frame don't co-mingle. And you have to keep track of the field order. Even numbered scanlines need to remain even, odd numbered scanlines need to remain odd. If not, you have reversed the field order and must encode accordingly.

I do all that in Virtualdub? Where? How?

You just have to be careful about what filters you use and when. For example, you can't blur the frame while it's interlaced. Some filter have an explicit interlace option (Resize, for example). Some filter can be used on either interlaced or progressive frames (brightness/contrast, for example).

As far as interlaced status identification, per your advise, I did advance through the frames using the "Bob Doubler - TFF" filter and, like I said, to me every frame was different from the previous, not 2 frames were near the same. With BFF though, it gave me a back and forth movement.

Because not 2 frames are alike, I would have thought I have a progressive video.

The fact that every frame is different after the bob double filter means that the video is interlaced. All scanlines of a progressive frame comes from a single picture, a single moment in time. For example, every second of a 30 fps progressive video has 30 different pictures, taken at 1/30 intervals. An interlaced frame contains two half-pictures (called fields), taken at two different times. So every second of a 30 fps interlaced video has 60 different half-pictures, taken at 1/60 second intervals. Think of it like this: the camera takes two different progressive 720x480 pictures at 1/60 second intervals. Then it takes half the scanlines (240) from the first picture, and half the scanlines (240) from the second picture, and weaves them into one 480 scanline interlaced picture before saving them. It repeats that 30 times a second to make a video.

If I have several capture pieces from the same tape is it best to append them in VirtualDub before processing or after? If before, should I make a new "raw" file containing all pieces and then process this file?

That's a personal choice and may depend on what other software you are using. In general, I would keep the project as several smaller files while working.

[drgt]

Should this be extracted in this folder?

[jagabo]

http://avisynth.org/mediawiki/SuperEQ

SuperEq("C:\Equalizer Presets\Loudness.feq")

Put them wherever you want. Use the full path in the SuperEq() call.