TESTING image sizes again. TIF upload for you to play with in Post #12.

[Andrea B.]

Thanks! I'll try my diff again.

 

But I haven't seen an increase in file size. I've seen decreases.

 

Added 31 July 2023: Uploaded JPG filesize can increase or decrease.

[Andrea B.]

I pixel peeped in PS at 3200%. I don't see signs of compression in the uploaded version. I do see slightly different per pixel colors. This could be due to a slightly different version of srgb used by the GB image app? But I don't see multiple pixels "merged" into one color like you would see in lower quality JPGs.

 

Bear with me here. I'm trying to find out what GD actually does.

 

Original image at 3200% enlargement.

Website image at 3200% enlargement. (Website image was dragged off website onto desktop.)

 

[Andrea B.]

Here is the Difference after applying Auto Contrast. I didn't know that trick about Auto Contrast to bring out subtle differences!!!

The differences seem mostly along edges. Why is that? Could this indicate some difference between how the image is sharpened (or not) rather than JPG quality change? (answer below:  no.)

[Andy Perrin]

JPG artifacts are generally along edges because of the way JPG compression works. That will always be the case. If the difference is caused by sharpening it will not have a blocky appearance. Blocks are caused by the JPG algorithm compressing 8x8 squares of pixels. To be able to distinguish you must examine the picture at the level where you can see the (potential) 8x8 blocks. Zoomed out, one cannot tell. 

[Andrea B.]

Makes sense then, what we are seeing in the diffs.

 

I looked last night at two JPGs - one saved at low quality and one saved at high quality. In that comparison I could definitely find areas of compression. In the example I posted above it was not so easy to find. 

 

I'm still amused that the black diff actually does have some differences if Auto Contrast is applied. I'm happy to have learned that trick.

*******

 

I do have the GD quality setting at 100, the highest. And I have the ImageMagick setting also at 100, although currently ImageMagick is not being used. So, the question becomes: if I upload a JPG of highest quality, why is its quality reduced by GD given the GD 100 setting? Also, why is the uploaded JPG rewritten/resaved at all? I'm thinking that happens because the GD "wrapper" is needed in order for dynamic resizing to be used on post and black box images. I'm going to visit the Invision forum and ask these questions. 

 

Another question:  do changes happen to PNGs? An easy test to make.

[Andrea B.]

PNG test.

First note that the PNG quality is set to the highest value 100 in our software.

I saved a raw Nikon NEF as a TIFF in Photo Mechanic.

I then saved the TIFF as a PNG with no compression in Photoshop Elements.

No other edits were made.

 

Original PNG dimensions: 4912 x 7360

Original PNG filesize: 103.5 MB

That is way too large for a fast upload, but I want a big one for test purposes.

The extracted JPG from the NEF file is only 2.0 MB.

 

Here's the uploaded PNG.

 

 

 

 

[Andrea B.]

I downloaded the website PNG and compared to the original PNG in a PS difference layer. I then applied Auto Contrast.

 

There were no differences between the uploaded PNG and the website PNG.

 

However, that uploaded file was enormous in weight. So I need to perform the experiment with a resized PNG. And also with a PNG saved with compression.

[photoni]

Andrea don't waste time with PNG or TIF
the web is JPG only

.
I understood that it is wrong to post JPG with 12/12 quality because the soft rewrites the file with about 8/12 quality
VISUALLY they are 99% similar but the UVP file weighs less

.
it is even more wrong to post a JPG file with 3/12 quality, because the soft rewrites the file with about 8/12 quality

VISUALLY they are 99% similar but the UVP file weighs much more.

[Andrea B.]

PNG Test

A raw NEF was saved as a resized TIFF in Photo Mechanic.

TIFF:  534 x 800 px with weight 1.4 MB.

The TIFF was resaved as 1) an uncompressed PNG and 2) a compressed PNG in PS Elements.

 

Uncompressed, resized PNG

534 x 800 px with weight 1.2 MB

 

 

Compressed, resized PNG

534 x 800 px with weight 450 KB 

[Andrea B.]

Now, I will compare those two uploaded PNGs to the original resized PNG.

 

Original small uncompressed PNG with website version:  NO differences.

 

Original small compressed PNG with website version:  NO differences.

 

 

Comparison of original small UNcompressed PNG with original small compressed PNG:  No differences.

 

 

---

 

BTW, if you read up on PNGs, you would already know the outcome of the 3rd test. But sometimes it is good to try to "prove" things yourself, yes?

PNG references:

PNG on Wikipedia

JPG vs PNG by Adobe

 

---

NOW, the question is this:  why does the forum software change a JPG but does NOT change a PNG?

I'll ask that question on the Invision forum and report back.

The second question is this:  For a resized file, what's the load hit on JPG vs PNG?

[Andrea B.]

Toni, I added clarifications to my original comments about JPG resizing to reflect that I now know how to "see" the differences between an original JPG and the uploaded, website version. This is thanks to Andy's showing me the difference trick of applying Auto-Contrast. 

 

I remain curious about using PNGs so I'm going to continue to look at that.

 

[Andrea B.]

1100 px width vs 1200 px width

Checking to verify that 1100 px is the display limit with a post in a fully expanded browser.

 

 

[Andrea B.]

The actual width limit within a post is 1093. Weird. I'm trying to find out how to expand the display area.

[Andrea B.]

retesting lightbox.

1600, 1700, 1800, 1900, 2000, 2100

 

 

 

 

[Andrea B.]

JPG limits seem to be:

1. 1093 px width within post (Really soooo weird. I'm trying to find where to change that.)
2. 1800 px width within black box if browser not expanded.
3. 2000 px width within black box if browser expanded with no sidebars.

I have not tested to determine whether any of those are system dependent.

 

Click-up limit off black box to view image on blank browser page is very likely system dependent. For Safari on a MacBook Retina, 3100 px width seems to be maximal before scrolling sets in. YMMV.

 

From what I know currently, I will probably recommend a maximum 1000 - 1800 px width. If properly resized prior to posting, that would give both a nice view within a post and within the black box without having to force the viewer to go through browser expansion and sidebar closure.

[Andrea B.]

Still waiting to hear from Invision about changes to uploaded JPGs.

[Andrea B.]

The Invision folks are seeing this now. No answers yet.

I'm going to try the diff test for a lower quality jpg than what I was using above. Then give them the diff to look at.

[Andrea B.]

Image cropped and resized (374 x 800) in Photo Mechanic. Saved as JPG quality 50/100.

[Andrea B.]

The auto-contrasted Difference between the original (cropped, resized) JPGq100 and the uploaded, website version.

 

 

 

The auto-contrasted DIfference between the original (cropped, resized) JPG/Q100 and a version saved at Q50. The 2nd Q50 version was made from a cropped, resized TIF master file so as not to incur any additional degradation due to resaving a changed JPG.

 

 

[Andrea B.]

Here is the auto-contrasted Difference between the original (resized, cropped) JPG/Q100 and a version for which Detail 10 (local contrast) was applied in Photo Ninja. The 2nd Detail 10 version was made from a cropped, resized TIF master file so as not to incur any additional degradation due to resaving a changed JPG.

[Andrea B.]

Here is the auto-contrasted Difference between the original (cropped, resized) JPG/Q100 and a version which was sharpened in Photo Ninja. The 2nd sharpened version was made from a cropped, resized TIF master file so as not to incur any additional degradation due to resaving a changed JPG. Sharpening strength was set to 75 with a radius of .50. (I do not know how PN sharpening compares to the PhotoShop USM tool.)

[Andrea B.]

Here is the auto-contrasted Difference between the original (cropped, resized) JPG/Q100 and a version which was saved at Q80. The 2nd Q80 was made from a cropped, resized TIF master file so as not to incur any additional degradation due to resaving a changed JPG.

 

[Andrea B.]

Compare the Upload/Download difference to either of the Quality diffs. The Upload/Download difference does not appear to me to  like the quality diffs. Click the first one up into the Black Box and then you can arrow through them easily. So what the heck is GD doing to our uploads?

[Andrea B.]

I thought I should add a difference between the JPG at highest quality Q100 and one saved at very low quality Q20 (out of 100).

Here is the auto-contrasted Difference between the original (cropped, resized) JPG/Q100 and a version which was saved at Q20. The 2nd Q20 was made from a cropped, resized TIF master file so as not to incur any additional degradation due to resaving a changed JPG.

[Andrea B.]

Whatever it is that GD is doing to our uploaded JPGs, it appears to be more like some kind of edge change than the typical JPG quality change???

Looks kinda like a Find Edges in that upload/download diff.