[1.11.5] 16:10 (8:5) and "golden ratio" aspect ratios grouped incorrectly in Catalog Filter > Properties > Aspect ratio

[KLE-France]

Ante-scriptum: I present below another case of specific aspect ratios, but, I do wonder if the entirety of aspect ratio calculations needs to be revisited. I speak below of a new, mysterious 14:9 aspect ratio appearing in 1.11.5, but there are others as well, e.g., a "9:7" which I had never seen before either... OR, this IS normal as part of a new-founded way to do things? IDK.


See my earlier post, currently in "Closed/Resolved", first. I think everything discussed there is pertinent here as well.

XnView: MP 1.11.5 - 64 bit (this issue did not exist at version 1.11.2)
OS: Windows 10 Home - 64 bit

In the previous post linked-to above, the problem was with 7:6 and 6:5 ratios wrongly categorized as 6:5 or 5:4 respectively.

Here, I report a new issue, that is, "golden ratio" (1:1.618) aspect ratios are grouped as 16:10 aspect ratios (now reported by XnView as 8:5; I will use this hereafter), and 8:5 ratios are now categorized as 14:9 (a new ratio I had not seen in earlier versions).

I suspect that the golden ratio images will just need to live in the 8:5 category, and that's OK. I'm stupid fussy about my cropping; using both 8:5s AND golden ratios is absurd on my part, but grouping them both as 8:5s is perfectly reasonable.

So, the following only looks at the 8:5s incorrectly categorized as 14:9s.

I sampled 20 recent images now landing in the 14:9 category. I know that I cropped all of them using Affinity Photo 2's crop tool set to a 8:5 aspect ratio (but 16:10 in the following images). 13 of them had an additional row of pixels included because the crop tool, set to an obligatory 8:5, lands on a partial row of pixels.


Here's an example of what that looks like (all images are screen shots, 8:5 (16:10) crop tool activated (the white lines), highly zoomed in on a corner of the image):

greater than 8-5_01.jpg

To not leave alpha on the edge of the image, the program keeps all of the partially covered row. Thus, for these 13 images, the problem might be due to a lack of rounding when the image is slightly larger than exactly 8:5 (see JLM's observations in the earlier post).



However, and perhaps more troubling, there were 7 images where all four edges of the crop tool fall on exact and complete pixel rows:


Here's the upper-right corner of one such image:

exactly 8-5_UR.jpg

and here's the lower-left corner of that same image:

exactly 8-5_LL.jpg

This seems more problematic. It would seem to me that the 7 images with "perfect" (no partial pixel lines) 8:5 crops should be in the 8:5 category...

Note too that all of the above screen captures were cropped using that same crop tool set to 16:10 (8:5).

Thanks and good day to all!

[xnview]

do you have sample files?

[jkm]

KLE, I agree with your sentiments, and it’s ok to be “fussy” about aspect ratios.

But arguments like this (and similar in your previous thread):

I sampled 20 recent images now landing in the 14:9 category. I know that I cropped all of them using Affinity Photo 2's crop tool set to a 8:5 aspect ratio (but 16:10 in the following images). 13 of them had an additional row of pixels included because the crop tool, set to an obligatory 8:5, lands on a partial row of pixels.

have no value and are beside the point. How Affinity Photo or any other app does its cropping is of absolutely no interest or relevance. In short, who cares.

The ONLY thing that matters is pixel dimensions and mathematical ratios, so best to keep this on point:

ALL aspect ratios (like 16:10 or 16:9) have a numerical value. For 16:10 it is 1.6000; for 16:9 it is 1.7778 approximately. Those two ratios differ by 10-11% (depending on how you calculate).

The only correct way for the app to group by aspect ratio is to calculate the numerical value of the ratio, and allow a certain set percentage tolerance range, like +/- 3%.

So for example, 16:10 +/- 3% is 1.552 to 1.648, and any image with a numerical aspect ratio in that range would be grouped as 16:10 (or 10:16 depending on orientation).

If two aspect ratios are closer together than the standard tolerance, then a choice must be made: either the standard tolerance must be lowered, or one of the ratios must be excluded.

It’s that simple.

There will always be images that are 1 pixel too many or too few to meet the definition. That’s math.

I will point out that the inclusion of the "Golden Ratio" is potentially problematic, as it is only about 1% away from the VESA standard 16:10. Inclusion of the Golden Ratio will at best cannibalize many 16:10 images, and at worst, lower the tolerance so much (1% instead of 2% or 3%) that many many more images will not be readily grouped. My advice would be to just let "Golden Ratio" images be lumped in with the 16:10 standard.

What I think is NOT correct, is what the app was doing in an example from the previous thread: an asymmetrical tolerance, like -2% / +0%

So I don’t think sample files should be needed…

Pierre should ensure the calculation is being done in the correct way, share what the tolerance range is, and that’s it.

Edit: In case Pierre wants it, here is a table of common aspect ratios, sorted by decimal, in both comma and space separated format.
The ones with decimals in the ratio, like 2.39:1, are cinematic aspect ratios (films/movies).

I’d not suggesting all these be explicitly listed. If I recall correctly, when the Aspect Ratio property tree was first created, we just defined all ratios larger than a certain value as “Panoramic”.

Code: Select all

Aspect Ratio,Decimal
1:1.618                        0.0618
9:21                           0.43
9:16                           0.56
10:16                          0.63
2:3                            0.67
5:7                            0.71
1:1.14                         0.71
3:4                            0.75
8.5:11                         0.77
11:14                          0.79
4:5                            0.8
8:10                           0.8
5:6                            0.83
1:1                            1
1.19:1                         1.19
6:5                            1.2
5:4                            1.25
14:11                          1.27
11:8.5                         1.29
4:3                            1.33
1.35:1                         1.35
1.37:1                         1.37
7:5                            1.4
1.41:1 (sqrt(2):1 or A series) 1.41
3:2                            1.5
16:10                          1.6
1.618:1                        1.618
1.66:1                         1.66
1.75:1                         1.75
16:9                           1.78
1.85:1                         1.85
2:1                            2
2.20:1                         2.2
21:9                           2.33
2.35:1                         2.35
2.39:1                         2.39
2.40:1                         2.4
2.55:1                         2.55
2.76:1                         2.76
32:9                           3.56
4:1                            4


Aspect Ratio,Decimal
1:1.618,0.0618
9:21,0.43
9:16,0.56
10:16,0.63
2:3,0.67
5:7,0.71
1:1.14,0.71
3:4,0.75
8.5:11,0.77
11:14,0.79
4:5,0.8
8:10,0.8
5:6,0.83
1:1,1
1.19:1,1.19
6:5,1.2
5:4,1.25
14:11,1.27
11:8.5,1.29
4:3,1.33
1.35:1,1.35
1.37:1,1.37
7:5,1.4
1.41:1 (sqrt(2):1 or A series),1.41
3:2,1.5
16:10,1.6
1.618:1,1.618
1.66:1,1.66
1.75:1,1.75
16:9,1.78
1.85:1,1.85
2:1,2
2.20:1,2.2
21:9,2.33
2.35:1,2.35
2.39:1,2.39
2.40:1,2.4
2.55:1,2.55
2.76:1,2.76
32:9,3.56
4:1,4