Convolution3D makes my image better looking but also noticeable darker ... why?

[FulciLives]

Hello :)

I have an NTSC (704x480) AVI capture from a VHS video (good quality pre-record) that was done with the PICVideo MJPEG codec (19 setting).

The capture turned out SUPER great but here is the problem.

I am using an AviSynth script because I'm using decomb for ITVC and then applying Convolusion3D using the movieLQ pre-set.

This is giving me a great quality progressive image BUT the image is noticeable DARKER than the original.

To double check this I ran 2 instances of VirtualDubMod and in one I loaded the AVI directly and in the other I loaded the AviSynth AVS file which of course is applying the Convolusion3D filter. No doubt when I do a frame by frame match-up and click back and forth between the two ... the AVS script version is darker and it is because of Convolusion3D

Is there a setting to change this or something else I can do to "combat" this?

Thank you in advance for any comments :)

- John "FulciLives" Coleman

[alucard83]

from what i can tell, just use vdub's built in brightness/contrast filter to help you clear that up

[stickboy]

Originally posted by FulciLives
To double check this I ran 2 instances of VirtualDubMod and in one I loaded the AVI directly and in the other I loaded the AviSynth AVS file which of course is applying the Convolusion3D filter.What you're probably seeing is the difference between the luma range output by the PicVideo MJPEG codec in RGB mode versus YUY2 mode.

Load two different AviSynth scripts--one with Convolution3D and one without--and you shouldn't see a noticeable difference in brightness.

[FulciLives]

Originally posted by stickboy
What you're probably seeing is the difference between the luma range output by the PicVideo MJPEG codec in RGB mode versus YUY2 mode.

Load two different AviSynth scripts--one with Convolution3D and one without--and you shouldn't see a noticeable difference in brightness.
Well thank you because you were completely correct. I made a second AviSynth AVS script that was a copy of the original except of course I took out the Convolusion3D lines and yeah the two (both with and without) look the same in VirtualDubMod.

So now I have another question ... I use TMPGEnc and there is an option called:

OUTPUT YUV DATA AS BASIC YCbCr NOT CCIR601

As I understand it this option has something to do with the 0-255 range VS the 16-235 range. I was reading some threads about this over at DVDRHELP.COM and they seemed to suggest that with a NTSC TV capture this option should be CHECKED (the TMPGEnc NTSC DVD Template defaults to this option being UNCHECKED).

I ran a sample clip (the first 6 minutes of the movie) doing one encode with the option checked and one without. The one that had this option CHECKED looked substantially DARKER than when it was UNCHECKED.

Based on that I am assuming it should be left UNCHECKED but any input would be appreciated :)

Again this was an NTSC VHS capture at 704x480 using the PICVideo MJPEG codec on the 19 setting (YUY2) and I'm using a fairly simple AviSynth AVS script that basically just uses decomb (for IVTC) and Convolusion3D and nothing else.

Thanks again!

- John "FulciLives" Coleman

*** EDIT ***
I just realized that I've been spelling Convolution3D incorrectly not only in the THREAD name but throughout my postings here. Sorry about that but I just wanted to say that for the record it is spelled correctly in my AviSynth AVS script :)

[Wilbert]

I ran a sample clip (the first 6 minutes of the movie) doing one encode with the option checked and one without. The one that had this option CHECKED looked substantially DARKER than when it was UNCHECKED.

Based on that I am assuming it should be left UNCHECKED but any input would be appreciated
You are correct. If you check it your luma will be scaled to [0,255].