The End NR4 - Stable - High - Intra - GH2 Patches - 03/28/2015 - by Apefos

apefos

one more attempt for the gop2 pal:

JayUKB

Is this, predominantly, for HBR (PAL)?

DouglasHorn

Just curious @apefos ... which version are you using for your cameras?

frullaccia

Gop 2 works ONLY in The End versions. The gt2 has a good hbr (best gop table, I suppose) and 24h. The standard version has got the best 24h (gop table), but shows a bad performance in hbr (noisy and artifacted). So: 24h in the standard version; hbr in the gt2.

Gop 1. The last group of patches has an improved hbr set at 120datarate (but I still prefer The End hbr gop2 gt2 or some gop3 like Climax or Epilog, the last one being very fine). As for 24h 120datarate, I would choose this last one, or the “old” version which can be found in your 400kb file. I also suggest to avoid the Gop1 versions from the gop1&gop2 zip file.

icp

@apefos I've tested the following patches against Valkyrie:

The End All-I (Intra GOP 1).
Epilog High120 (GOP 3 I believe).

PAL, natural light, interior, manual lenses (MD Rokkor-X 28mm 2.8), all electronic features off, smooth profile all at -2), iso 160, 320, 640 and 1250, 1/50 and 2.8 aperture. I've only tested 24p H.

At 100% both are great, I have to zoom to 400% to check differences.

The End All-I Gop 1 is beautiful, on pair with Valkyrie. No mud or macroblocking anywhere. Grain "jumps/dance" a little more than Valkyrie.

Epilog isn't so good to my eyes. Grain is slighty "square" and "dance" a lot more, and I see mud and macroblocks in shadows and blacks.

For me: All-I is a top patch. Awesome work. I guess is just a matter of taste, but I prefer to have organic grain, than artifacts from denoising.

frullaccia

@apefos For me, The End is the best stable 120datarate patch. But I like the different picture of Climax and Epilog.

apefos

There is a hope for the PAL GOP2, I am doing some tests with various gop tables.

apefos

THE END NR4 - ALL GOP VERSIONS - ALL DATARATE VERSIONS

in the patch name you will find five numbers for gop, they are in this order: 24p, 25p, 30p, 50p, 60p

example: GOP 1-1-3-2-4 = 24p gop1, 25p gop1, 30p gop3, 50p gop 2, 60p gop 4

this way you can find the gop you want, including intraframe and the new gop 2 working ok.

Tip: if you like the other patches (catharsis, climax, denouement, epilog, the end nr1, nr2, nr3, nr5, nr6) you can use these ini files below as a base and copy/paste the matrices from the patch you like to the ini files uploaded below.

If you want to copy/paste the matrices from other developers to these ini files below maybe you will you need to change the deblocking tables. To avoid the diagonal rain pulsing pattern when using matrices from other developers try the deblocking tables as they are, if you get rain or instability you can try to use this number in all deblocking table positions: 0x418 or this: 0x410 or this: 0x420 (just when using matrices from other developers)

A good idea to clean the diagonal rain and improving stability when using matrices from other developers in the ini files below is to round all numbers in the matrices and in the deblocking tables to multiples of 3 or to multiples of 2. Small numbers round up, big numbers round down. Similar multiples in the deblocking tables. Using a common denomitator 2 or a common denominator 3 in the matrices and in the deblocking tables helps stability and helps to avoid the diagonal rain.

if you are a patch developer feel free to use the ini files below as a base to implement your matrices or to develop new matrices.

second upload is correct

apefos

the second upload in previous post is correct.

MaartenL

I'm a little lost. I want to shoot 25p with my SanDisk Extreme PRO SDXC UHS-I 64GB 95MB/s cards. My record times are up to 15minutes each, but most of them <2 minutes. Which patch should I take? :)

apefos

@MaartenL in the last upload all patches works ok, the choice is more related to the look you want in noise and texture. So it is good idea you try and keep in camera the one you like most.

If you will use various auto electronic features at the same time then the stable version is recommended. If you use just one or two auto electronic features at a time all versions can work good. Auto electronic features are: Auto Exposure, Auto Focus, OIS image stabilization, iDynamic, iResolution, Noise Reduction. (noise reduction disabled in -2 is better image quality, in the film mode) This is more related to 720p, in 24p these features can work together in all versions.

Native lenses uses some camera processor also, using manual lenses with adapters allow the camera processor to be more free to do the image encoding. All my tests was done with the 14-42mm kit lens so all versions are working with this native lens.

But no need to worry, the camera will not freeze, no need battery pull. if the camera processor cannot handle various auto electronic features at a time, then the camera stop recording, saves the file and go back to stand by mode. The stable version was developed to avoid stop recording with the best possible image quality under safe stability, the other versions was developed to get a better image quality, placing stability in the limit.

The different GOP in the versions delivers different texture and noise, looking close to the screen you can perceive. Some people like to see the image in slowmotion (half speed) and/or in 200% or 400% zoom in timeline to perceive the noise/texture better when choosing a patch.

MaartenL

Thanks apefos. I think I prefer c_the-end_nr4_high104_gop2-2-3-2-4. Anyway, it is hard to compare.

producer

@apefos: Going to try the latest ALL-GOP patches, but below is my graded test of "THE END NR4 INTRA 120". Not sure if it's because of the grading, but seems to have much more noise than my previous tests above with the older patches, even that here I used ISO4000.

Also I make a really heavy grading so that to reach some more density and color shifts of GH2, trying to avoid its "thin" picture.

apefos

The two versions below are the gop2 versions with the intraframe matrix.

Compare seti below vs. setc from previous upload, compare setj below vs. setd from previous upload.

frullaccia

@apefos Relating to @producer's post, among the different versions of The End, it seems that there are fine options choosing one of the previous. This is also true for some of 120datarate patches, like Epilog, which I use in a "old" version (I think it is the one coming from the 27 march Apefos post), and only at 24h (because that version is not working in hbr).

Anyway, this last (1 April) matrix of Gop2 is the most efficient (shadows).

At this point I am very uncertain about which one is the best patch.

Gop2 is a very promising choice; in old patches the picture was excellent but not fluid; now it is fluid, even if maybe a little less interesting. I suppose this Gop is very difficult to be modified.

At 120 24h I am using two "old" versions of Epilog and The End.

Gop3 performance is remarkable in versions like, for instance, The End 30march; Epilog (27march?); and Climax.

Talking about non Apefos patches, Valkyrie is a good one, worthy to apply the great Apefos pulsing pattern correction.

apefos

@frullaccia I am in the need of advance some tasks which are late due to my work in patches. If you or someone else are willing to help, you can copy paste the matrices from the other patches to these ini files below. these ini files below have all improvements until now.

small tutorial: open both ini files in notepad text editor. right click/open with/notepad. or you can rename the ini files to txt and double click on it (seta.ini to seta.txt). and after edit the file rename again (seta.txt to seta.ini). then you copy the matrices from one patch and paste on another patch file and save as. in the first lines of patch file you will see the patch name, you can change it, no problem. this is the text which appears on the yellow box in ptool.

to understand how to copy/paste the matrices, you can read and learn what are the matrices numbers in the setj.txt file uploaded here.

you can copy/paste matrices from catharsis, climax, denouement, epilog, the end nr12356 to these ini files here. no need to change deblocking tables.

you can upload your copy/paste patches, so other people can try.

if you or someone else are willing to tweak matrices from other developers, it would be good idea to talk to the developers or to someone who understand matrices and ask them to tweak the matrices using the tips below:

If you want to copy/paste the matrices from other developers to these ini files below maybe you will you need to change the deblocking tables. To avoid the diagonal rain pulsing pattern when using matrices from other developers try the deblocking tables as they are, if you get rain or instability you can try to use this number in all deblocking table positions: 0x418 or this: 0x410 or this: 0x420 (just when using matrices from other developers which have common denominator 2). the deblocking tables digits must have the same common denominator from the main matrices digits.

A good idea to clean the diagonal rain and improving stability when using matrices from other developers in the ini files below is to round all numbers in the matrices and in the deblocking tables to multiples of 3 or to multiples of 2. Small numbers round up, big numbers round down. Similar multiples in the deblocking tables. Using a common denomitator 2 or a common denominator 3 in the matrices and in the deblocking tables helps stability and helps to avoid the diagonal rain.

to understant what are the deblocking tables you can read and learn in the seti.txt file below.

Sorry but I cannot do, I have lots of other things to do now.

It is time now for the community to show it's power. it is time for the other developers tweak their matrices, to round the numbers and avoid the rain. the ini files below can handle all kind of matrices, they are perfect base for all matrices from other developers.

frullaccia

@apefos In this period, time is my enemy. But as soon as I will change-combine for me some of your patches, I will share it on this topic with the personal view community. Thank you!

frullaccia

@apefos and friends. I have tried to change a little your last Gop2 patch. I am sure that I made something wrong, but this experiment seems to be working and I seriously doubt it would with a Gop1 patch.

The picture of this The End Alpha Gop2 version is stable and clear.

I noticed that I prefer your low quantization patches, so I just tried to modify a little some parameters from your tables. I really didn't study the quantization process. Anyway the result looks good to me.

While making this I was tricking through different settings. I'd say this is a try, but a reasonable one.

producer

@apefos: A quick test of "THE END NR4 HIGH 104 GOP 2-2-2-2-2 M36" with ISO2000 and different GH2 WB and film modes - cinema, smooth, vibrant, cinema:

apefos

@frullaccia I am happy to see you are doing some try. So I developed a new matrix, apefos 444, inspired on Cinema Smooth from Valkyrie and Intravenous, for you to copy/paste. The txt file below has the apefos 444 new main matrix, new fallback, and new deblocking table. you can use the end patches a, b, c, d, e, f, g, h, i, j as a base for this. copy from txt and paste on the ini files. I hope it is artifacts, moire, aliasing, rain free... I used all my knowledge on it.

I am very busy now, so ask for help from other P-V members to help you understand things, there are lots of people here who knows how to copy/paste settings. You can find many people in all the topics in the Top Settings section of P-V, including my old topics.

Good Luck!

producer

Another day indoor low light test at ISO1250 with "THE END NR4 HIGH 104 GOP 2-2-2-2-2 M36":

apefos

apefos_444_02

another version with different chroma matrix and chroma fallback for copy/paste.

haribabis

@producer, you are killing me! Have you made any post on the procedure of the color grading you make ? i am very interesting in such a post!

producer

@haribabis: All you have to start with is to use cinema profile (even with contrast 0 or +2), no matter what some users say it's oversaturated in highlights or crushing the blacks, etc. Actually, cinema profile is what can be seen in so many movie productions as an image characteristics.

@apefos: Just a suggestion about an experiment - is it possible for you to modify the matrix so that the macro (coarse) sharpness is increased instead on the micro (thin, micro) sharpness? Hope you get what I mean.

apefos

These are attempts to make the image more sharp, maybe they are not so good for high iso...

copy and paste the matrices using "The End NR4" patches as a base.

also copy and paste the deblocking table to all deblocking flag and high.

Howdy, Stranger!

Categories

Tags in Topic

Top Posters