Wolverine-Hawkeye Telecine

[David Brown]

Kamel here is the Alibaba link. I have never purchased anything from them before. The specs of this lens is very difficult to find with mainstream sellers.
https://cnaico.en.alibaba.com/produc...oard_lens.html
The lens number is an Aico ACH1256MAC I hope that means 12mm, f5.6, macro!

Stan seems to have found a lens of interest. I'll wait with interest for that.

I have found inconsistencies even inside the same companies with reported specifications of different lenses. No consistent template of features or performance exists. Why do most lenses have a working or minimum focus at around 0.1 meter whether they are macro or not? Surely that distance can be reduced by increasing the lens to sensor distance?

I have ordered 2 cheap 16mm lenses, one "made" for 1/2.5" and another for 1/2". Can't wait to see the "circle diameter" projected from each at the same distance.

Lens calculations can be daunting when all the input criteria cannot be identified. "Even then your actual mileage may differ"!!!

[Stan Jelavic]

This lens might work. We would be somewhere between 32mm and 40mm film distance from the lens. The test picture of sugar cube shown was taken at 50mm and looks very clear. F stop of 1/5.6 should give pretty good DOF.
https://www.scorpionvision.co.uk/m12...5-6-macro-lens
No IR. Could potentially have some impact on contrast. To be tested. Possibly add IR block on top of the gate if required.

Not an expert on macro lenses David but my understanding is that macro lenses are designed to work at close range to give you high magnification.
That does not mean that a non macro lens cannot give you that same magnification if the working distances are the same but macro will be sharper
because it is designed for close range not all the way to infinity.
My assumption is once we push closer than the specs, the macro will be sharper. My guess. Have not researched it.

[Bruce Davis]

Hi Stan Just had a chance to test your VideoFred script (Hawkeye), have run into a problem:-
"StackHorizontal Image Heights Don't Match Line 271"
I assume it has something to do with W2= and H2=, this was previously calculated and in your script you have manually entered numbers.
How does one calculate the correct numbers (if this is the problem). My video is 1344 x 968 and have tried setting W=720 H=576 also W=1920 H=1080, not sure how to overcome this problem.
Thanks for any help you can give
Regards - Bruce

[Kamel Ikhlef]

Thanks David,
I think your 2/3" M12 lens choice is good for the imaging source sensor (1/2.5").

[Bruce Davis]

Thanks David, Kamel and Stan for your efforts in this project, I do not understand anything about these lenses but it would be nice to find a good lens that covers all or most of the sensor area with the gate area. Not sure if its possible to have incorporated into such a small lens an aperture adjustment.
Thanks guys - Bruce

[Stan Jelavic]

Hi Bruce.
In Fred's script he reduces the final size to 720 by 576 because of interleaving.
I did not like that and bypassed interleaving and have the final size set to the original minus the crop factor on all 4 sides.
The crop is 32 pixels on each side so your output should be the original minus 64 for the height and the width.
So in your case you can hardcode:

W2= 1280
H2= 904

That should work.

Here is the code that sets the crop. I will fix the script to set the final size automatically.

#SIZE, CROP AND BORDERS PARAMETERS
#----------------------------------------------------------------------------------------------------------------------------
CLeft=32 CTop=32 CRight=32 CBottom=32 #crop values after Depan and before final resizing

[Bruce Davis]

Thanks Stan that worked.
Hi Stan, just did a comparison using the "IC Capture Settings" in your guide.
Using the Scripts, your Hawkeye default script (result8) and VideoFreds using Result4.
Gamma was set to 1.3 in both scripts.
VideoFreds Result4 script colours were set at blue- 1.0 and red= -2.5.
Other settings were the same.

[Bruce Davis]

The only difference that I can see is that Hawkeye Script has a little more red.

[Stan Jelavic]

Hi Bruce. When I compare I get different color of the sky. Hawkeye is more like Wolverine because I tuned it like that. Also Hawkeye is full resolution.
Option4 is 1/2 resolution.

[Bruce Davis]

Hi Stan, thanks for the reply, your results are slightly different to mine (to be expected) but I do not understand how you come to the conclusion that using "result=4" would halve the resolution of the image (or frame). one should be able to see a dramatic difference in image detail.

Your added "Hawkeye Script" image is more grainy than the "VideoFred script" image (as is mine, sample below), not sure why that is the case as the settings are the same, unless it does have to do with half resolution. Why would VideoFred write a script that halves resolution, his script is meant to improve not degrade.


Regards - Bruce

[Bruce Davis]

Thought about it afterwards to include all 3 images

[Stan Jelavic]

Certainly good questions Bruce. I do not have all of the answers for you but do have some.

VideoFred resolution is fixed to
W=720 H=576 #final size after cropping

If you try to change this line to higher rez you will get bad artifacts caused by MFlowFps()
I bypassed this function and went to original resolution but my video is more grainy.

Now bypassing MFlowfps() may not be a good idea because there is some additional sharpening and smoothing there.

interpolated= denoised.MFlowFps(super, backward_vec, forward_vec, num=numerator, den= denumerator, ml=100)\
.sharpen(last_sharp,mmx=false).sharpen(last_sharp, mmx=false).blur(last_blur,mmx=false)

So I added just the sharpening part and smoothing without MFlowfps()

smoothed= denoised.sharpen(last_sharp,mmx=false).sharpen(las t_sharp,mmx=false).blur(last_blur,mmx=false)

And then call the smoothed output instead of interpolated in result8

result8= smoothed.coloryuv(off_U=blue,off_V=red).converttoR GB24().colorshift(blue_shift= bluephase_shift, red_shift= redphase_shift)\
.converttoYV12().addborders(bord_left, bord_top, bord_right, bord_bot).sharpen(last_sharp,mmx=false).sharpen(la st_sharp,mmx=false).blur(last_blur,mmx=false)

This gives me less noise but option4 is still better even if I increase the output size to match my script output.
https://photos.app.goo.gl/WSuUHDBckZWozRpNA

This needs some more work.

Here is the new Hawkeye script:
https://drive.google.com/open?id=1Lq...bsUvWkC0XLeOXS



Looks like other people had similar issues with MFlowfps() at higher resolution:
https://forum.doom9.org/showthread.php?t=165075

[Stan Jelavic]

More info on script issues:
https://forum.doom9.org/showthread.php?t=165075&page=1

[Bruce Davis]

Hi Stan, not being a programmer but having read the information from your link, if I understand this correctly, VideoFreds script has purposely reduced or scaled the resolution to 720x576 first to overcome the interpolation problems with mvtools "MFlowFPS" when the video is up scaled, to say, as in my case 1920x1080.

The flashSCAN8 Resolution is also only (3CCDs, horizontal 800 TV lines, Pixel 795H x 596V PAL, 811H x 508V NTSC) which is then up scaled to 1920x1080. Maybe as far as S8 film is concerned a higher resolution would not achieve greater detail from the home movie film.

If Film9 does not down scale to 720x576 from our captured resolution, for example 1256x924, before upscaling to 1920x1080 do we achieve any more detail? Does film9 use the same scripting DLL's as VideoFred and therefore the same methods (downscale first to 720x576)?

Thanks and regards - Bruce
P.S. would there be any improvement if we used the latest mvtools-2.7.39-with-depans.7z
http://avisynth.nl/index.php/MVTools

[Stan Jelavic]

Bruce, the script does not do any up scaling. Your final output is 720x596 if you use VideoFred script. The script scales any input video to this size and works with that. There is no change in scaling for the final output.
Actually, the script is very old - 2012 latest version? I do not see any updates since then, except for people adding enhancements but nothing is consolidated, only a patchwork. I would like to update it but it would be quite a bit of work.
I don't think it is only MFlowFPS that is broken with higher resolutions but also other filters like degrain. It may be to do with the vectors:

vectors= cleaned.MVAnalyseMulti(refframes=denoising_frames, pel=2, blksize=block_size, blksizev= block_size_v, overlap=block_over, idx=1)
denoised= cleaned.MVDegrainMulti(vectors, thSAD=denoising_strenght, SadMode=1, idx=2).unsharpmask(USM_sharp_ness3,USM_radi_us3,0)

Someone mentioned that the vectors are 8 bit only and that does not work for higher rez. We have to run these intermediate steps and compare the results. It is quite a bit of work.
Film 9 does not do much. It uses VirtualDub and its filters that can be turned on through a gui. It does not use FredVideo script.
When I use it I do not use any of the filters because they did not work well for me.

So in summary, we can try updating VideoFred script to get it to a reasonable quality level. One step at a time. The latest tools may also help. Will try that. Good idea.