Which picture better simulates the original?

[zweifingerjoe]

Hi folks, I have three pictures for you:

X: https://i.ibb.co/X41Z8rw/Original.png (https://ibb.co/X41Z8rw)
A: https://i.ibb.co/P53xh8s/a.png (https://ibb.co/P53xh8s)
B: https://i.ibb.co/m5NbzLM/b.png (https://ibb.co/m5NbzLM)

X is the original, A and B are reencodes. When you have to select between A and B, which one do you prefer with the original in mind?

Please note that you have to click on the picture again when you are at the image hoster's page to unlock the uncompressed png.

Please do not make assumptions about the codecs unless i revealed them after enough people had participated.

[poisondeathray]

A is more similar in terms of characteristics, preserving more of the original compression artifacts. From a compression standpoint it's "better"

B appears filtered (or the encoder effectively filtered it) , less similar, although general public would probably prefer it subjectively if you don't look at the original

[Emulgator]

Different resolutions make it harder to compare.
I would like to suggest giving a and b in 1920x1080, not 1920x1040
or give the cropped 1920x1040 that was the source.
I will go ahead and crop that Original myself.

After cropping: At first I go with poisondeathray.

P.S. Then after ~20 spot comparisons: Can not decide anymore.
Certain parts are kept differently in a vs.b.
Video will tell, maybe.

P.P.S: And now after ~10 more spot comparisions
I think that the knitware and the grain over skin is modeled nicely in B.
Curious to see the video and maybe revise myself again ;-)

[zweifingerjoe]

I would like to suggest giving a and b in 1920x1080, not 1920x1040
or give the cropped 1920x1040 that was the source.

If I crop the image its color appearance changes after uploading, there is something going wrong with the color management.

[Sharc]

B is sharper and preserves more details IMO than A. But the conclusion depends which parts of the picture I am looking at.

What's the purpose of the comparison? It is very questionable to qualify an encoder based on just a single frame. The conclusion for the next frame may be different.
Are all pictures of the same type, i.e. I, B, or P frames?

[zweifingerjoe]

What's the purpose of the comparison? It is very questionable to qualify an encoder based on just a single frame. The conclusion for the next frame may be different.
Are all pictures of the same type, i.e. I, B, or P frames?

I didn't verify the frame type, so it is likely that they are different. The purpose is that i want to know how people see the differences. Of course this is no scientific comparison, but it will be good enough for me as the characteristics of the differences seen in this particular frame are reproducible in other frame comparisons of the same movie. This is all through my individually biased perception of course.

tl;dr
I looked at different frames, the one I chose contains all the differences I personally noticed in every frame comparison. This is no scientific experiment.

[birdie]

I barely see any difference at 1080p on my 13.3" display even at 150% magnification.

[Greenhorn]

So, I wouldn't give my opinions as much weight as most anyone else here, but just for the heck of it:

Both seem to have lost some graininess compared to the source. I'd actually say B looks grainier and sharper, but that the grain looks a bit blotchy on the actor's cheeks in particular.

There's a big flat area on the actor's right lip in the source that's amplified in A but oddly much less noticeable in B.

Both of the images have different hue/lightness than the source-- that color management issue you were talking about, I guess.

All of these are subtle enough (to me) that I'm prone to second-guess them. Again, I'm not an expert.

[Blue_MiSfit]

I significantly prefer B.

The brightness of the image is different across these 3 tho...

[zweifingerjoe]

The only difference I found that could explain the color mapping issue is that the original is 8bit, but a and b are 16 bit. This is because a and b were encoded with 10bit so ffmpeg used 16bit color depth for the pngs. There is no color profile attached to neither of the images. On my (8-bit) monitor the brightness is the same over all three images.

======== original.png
ExifTool Version Number : 12.08
File Name : original.png
Directory : .
File Size : 3.8 MB
File Modification Date/Time : 2021:02:28 13:31:24+01:00
File Access Date/Time : 2021:02:28 13:31:24+01:00
File Inode Change Date/Time : 2021:02:28 14:01:03+01:00
File Permissions : rw-r--r--
File Type : PNG
File Type Extension : png
MIME Type : image/png
Image Width : 1920
Image Height : 1080
Bit Depth : 8
Color Type : RGB
Compression : Deflate/Inflate
Filter : Adaptive
Interlace : Noninterlaced
Pixels Per Unit X : 1
Pixels Per Unit Y : 1
Pixel Units : Unknown
White Point X : 0.3127
White Point Y : 0.329
Red X : 0.64
Red Y : 0.33
Green X : 0.3
Green Y : 0.6
Blue X : 0.15
Blue Y : 0.06
Gamma : 1.961
Image Size : 1920x1080
Megapixels : 2.1
======== a.png
ExifTool Version Number : 12.08
File Name : a.png
Directory : .
File Size : 10 MB
File Modification Date/Time : 2021:02:28 13:16:03+01:00
File Access Date/Time : 2021:02:28 13:15:40+01:00
File Inode Change Date/Time : 2021:02:28 14:01:03+01:00
File Permissions : rw-r--r--
File Type : PNG
File Type Extension : png
MIME Type : image/png
Image Width : 1920
Image Height : 1040
Bit Depth : 16
Color Type : RGB
Compression : Deflate/Inflate
Filter : Adaptive
Interlace : Noninterlaced
Pixels Per Unit X : 1
Pixels Per Unit Y : 1
Pixel Units : Unknown
White Point X : 0.3127
White Point Y : 0.329
Red X : 0.64
Red Y : 0.33
Green X : 0.3
Green Y : 0.6
Blue X : 0.15
Blue Y : 0.06
Gamma : 1.961
Image Size : 1920x1040
Megapixels : 2.0
======== b.png
ExifTool Version Number : 12.08
File Name : b.png
Directory : .
File Size : 11 MB
File Modification Date/Time : 2021:02:28 13:19:12+01:00
File Access Date/Time : 2021:02:28 13:19:12+01:00
File Inode Change Date/Time : 2021:02:28 14:01:03+01:00
File Permissions : rw-r--r--
File Type : PNG
File Type Extension : png
MIME Type : image/png
Image Width : 1920
Image Height : 1040
Bit Depth : 16
Color Type : RGB
Compression : Deflate/Inflate
Filter : Adaptive
Interlace : Noninterlaced
Pixels Per Unit X : 1
Pixels Per Unit Y : 1
Pixel Units : Unknown
White Point X : 0.3127
White Point Y : 0.329
Red X : 0.64
Red Y : 0.33
Green X : 0.3
Green Y : 0.6
Blue X : 0.15
Blue Y : 0.06
Gamma : 1.961
Image Size : 1920x1040
Megapixels : 2.0

[Zebulon84]

A seems smoother than the original,
B seems sharper than the original.

For a still image I prefer B, but for a video I would probably prefer A to avoid dancing noise introduced or increased by the sharpening.

[SeeMoreDigital]

@zweifingerjoe,

Out of interest...

What video format has your source been encoded with?
Why do you feel the need to process and generate a new encode?

[kolak]

1 frame is not a very good comparison. At least are they the same type: I,P,B?
Few frames in sequence is much better.
B looks better, but no one will see those differences during real watching scenario, so pick one which encodes faster.