VirtualDub causes odd deinterlace effect in Hybrid?
 

RE: Cropped captured AVI on VirtualDub causes odd deinterlace effect on Hybrid?

So I have captured an NTSC Video8 tape using VirtualDub with the HuffYUV codec. As I understand, it is captured at 720x480, which is a stretched 3:2 scale (SAR).

I've been following this guide (link) to crop the lossless capture in VirtualDub. After cropping, I render the newly "lossless" AVI and process it with Hybrid, applying only deinterlacing (QTGMC). However, I've noticed that the resulting mkv file appears wobbly. I suspect this may be due to incorrect cropping, despite following the guide's instructions to remove overscanning and headswitching by removing an even number of pixels on each side.

I'd really appreciate some guidance on this. Here are my settings and the video files:

https://drive.google.com/drive/folde...VB?usp=sharing

Hm, maybe it's due to a lack of a line TBC inside the camcorder you use? What camcorder model are you using. The rather shaky footage makes the wiggling difficult to pay attention to, but it is present.

I myself do cropping, deinterlacing and resizing all in Hybrid, if that would help make any sort of difference.

Looks like you're resizing before the deinterlace. That alters the geometry of each field (each line takes up more than 1px, but the deinterlacer still expects 1px per line), so you should hold off on that kind of thing til afterwards.

You cannot crop interlaced footage, and retain the original frame size. It must be masked/padded, which is what the guide shows.
https://www.digitalFAQ.com/forum/vid...erly-crop.html

Or resize after deinterlace.

Also, why deinterlace? Do not do it "just because", but for specific reasons.

Retain the original interlaced master, save a copy as deinterlaced.

The wobbling doesn't happen if I deinterlace an AVI file that is not cropped. I am using a Sony camcorder that doesn't have a TBC, however, I'm using a DMR-ES15 as a "TBC". this is my workflow:

Sony CCD-TRV12 →Panasonic DMR-ES15 → IO-DATA GV-USB2 → VirtualDub 1.9.11 → Hybrid

Would it be more effective to perform both the cropping and deinterlacing in Hybrid simultaneously, rather than cropping in VirtualDub before deinterlacing? It seems like I might be approaching the process in the wrong order.

If the guide shows how to mask the capture, wouldn't cropping be unnecessary? In the second step, it demonstrates covering the undesired areas of the picture, which aligns with my goal. How should I approach eliminating overscanning and headswitching in the video? Deinterlace first, then mask/crop?

I've kept the original capture untouched. I simply use that lossless file to generate a deinterlaced video through Hybrid so that I can stream it to my devices.

No. That's not how VirtualDub works. "masking" is technically the action of cropping + adding back black pixels.

Cannot pick steps, must follow all.

Why eliminate? What's wrong with just masking? You don't see either on your TV. The masking to black just saves bitrate and file size.

But yes, you can deinterlace first, crop second. But "should you" is the question to first ask yourself.

What devices?

I am a little confused. Do my attached images depict cropping or masking? I attempted to follow the guide provided step by step. This involved setting the resolution to 720 x 480 and adjusting the X1 offset to 12, X2 offset to 8, Y1 offset to 2, and Y2 offset to 6. These adjustments seemed to conceal overscanning and head-switching, but unfortunately, they led to some deinterlacing problems with Hybrid.


Smartphones, TVs, basically sharing them online via YouTube, Google Photos or messaging apps.

Given all the info, deinterlace first, crop + resize next, That easy here. :)

Is there a way to use QTGMC deinterlacer with VirtualDub? I've heard that using Yadif isn't as effective as QTGMC.

My goal is to handle all the cropping, resizing, and deinterlacing in the correct sequence within VirtualDub since it would keep the consistency of using the HuffYUV codec. With Smart Rendering, i noticed that it maintains a bitrate similar to the original capture, almost appearing lossless. I'm hesitant about switching between VirtualDub and Hybrid as it would entail re-encoding from HuffYUV to h.264. While I don't mind the extra storage space, I want to keep one AVI file untouched and create another that's cropped and deinterlaced. Then, I plan to process the latter with Hybrid for final distribution.

No QTGMC in VirtualDub, I wish.

However, Hybrid can output as FFHuffyuv (FFMPEG version), Ut Video, unofficial ProRes422, or FFV1. So it has lossless output options. (I need to see if he would consider adding original Huffyuv and/or Lagarith.)

For the sake of longevity, future codec obsolescence, I'd suggest the unofficial ProRes422, or next maybe Ut, from those choices. Not the FFHuffyuv, as it's not the same, and not as widely compatible.

So capture = Huffyuv archived master
ProRes422 (in .mov container) = deinterlaced master
And from that ProRes422, make the H.264 Or 265 in Hybrid.

With QTGMC, try "faster" preset first, not slowest. Slowest blurs.

Thank you for your guidance. I have successfully managed to solve my initial problem by doing the following process under your recommendations:

1. Capture tape with VirtualDub (with HuffYUV codec) [Archived master]
2. Deinterlace captured AVI with Hybrid (QTGMC faster) using ProRes codec. [Deinterlaced master]
3. Take .Mov and apply Crop/mask + Resize + PAR setting with Hybrid and encode to x264. [For sharing/streaming]

This gives the exact picture I was hoping to get from the start. I have two questions regarding the above:

• Firstly, when it comes to QTGMC deinterlacing with ProRes422, I'm debating between using the LT or normal profile. While ProRes is considered "visually lossless," I wonder if the LT profile might be more effective. My original AVI capture file has a bitrate of 63.3 Mb/s. When encoding with the LT profile, I get a bitrate of 58.5 Mb/s, compared to 84.9 Mb/s with the normal profile. It seems the LT profile mirrors the original bitrate more closely, despite being lower. Which profile do you recommend using?
  
• Secondly, in the third step mentioned above, where I'm cropping/masking in Hybrid, I'm removing 20 pixels horizontally and 8 pixels vertically to "hide" the headswitching and overscanning, resulting in a 700x472 resolution before resizing. Now, when I resize back to 704x480, does this constitute as upscaling or simply resizing? What exactly does this step do to the image? Could I unintentionally degrade picture quality by resizing it to a larger resolution than its original size?