image size in Photo

[KarinC]

Using resample when cropping works. Thank you @thomaso

 

edit: also using Lanczos 3 (non separable) under export helps, too. The image is still a little blurry using bilinear.

[thomaso]

28 minutes ago, KarinC said:

I was using unconstrained when cropping - is that the issue?

Probably. With "unconstraint" you may stretch the image and modify its aspect ratio which forces every pixel to different content.

[John Rostron]

18 hours ago, nickbatz said:

...because the screen resolution is what counts.

But resampling while you resize sharpens the image, and - correct me if I'm wrong - it never creates artifacts.

An appropriate downsizing algorithm would probably sharpen an image. However, I would be reluctant to claim that it never created artefacts.

If you are upsizing, then it is much more dependant on the kind of image - and on the algorithm. It  would also be more likely to cause artefacts because it has to create pixels which were not there before. Having said that, modern specialized upsizing software can do amazing things.

John

[nickbatz]

4 hours ago, John Rostron said:

An appropriate downsizing algorithm would probably sharpen an image. However, I would be reluctant to claim that it never created artefacts.

If you are upsizing, then it is much more dependant on the kind of image - and on the algorithm. It  would also be more likely to cause artefacts because it has to create pixels which were not there before. Having said that, modern specialized upsizing software can do amazing things.

John

Well, downsizing is going to sharpen things by definition, since everything is smaller!

But yeah, when enlarging the question is where the line between artifacts and acceptable results is.

Oh, and I should get in a dig while we're on the subject: Gigapixel AI is no longer in my good graces. It would be if I were taking pictures of birds, but for abstract modern art... failure.

[walt.farrell]

3 hours ago, nickbatz said:

Well, downsizing is going to sharpen things by definition, since everything is smaller!

Really? 

Generally when you downsize you're throwing away information (pixels). Why would the result be sharper?

Consider, for example, downsizing an image from 1000 x 1000 px to 500 x 500 px, or from 1,000,000 total pixels to 250,000 pixels or 1/4 of the data.

 

[RichardMH]

Not exactly sure what you are doing. You know you can resize in centimetres and keep the pixel count high by having a high dpi?

[nickbatz]

8 minutes ago, walt.farrell said:

Really? 

Generally when you downsize you're throwing away information (pixels). Why would the result be sharper?

Consider, for example, downsizing an image from 1000 x 1000 px to 500 x 500 px, or from 1,000,000 total pixels to 250,000 pixels or 1/4 of the data.

 

 

Everything is closer together = pixels are less spread out = everything is sharper = the earth is round.

[walt.farrell]

9 minutes ago, nickbatz said:

 

Everything is closer together = pixels are less spread out = everything is sharper = the earth is round.

f you kept the same number if pixels, and just changed the metadata to say the image was now 300DPI rather than 72 DPI, what you say makes sense.

But usually when someone says they downsized an image, they removed pixels.

[nickbatz]

36 minutes ago, walt.farrell said:

 

But usually when someone says they downsized an image, they removed pixels.

I just mean you're shrinking it. Of course removing pixels removes pixels!

[Alfred]

5 hours ago, nickbatz said:

I just mean you're shrinking it. Of course removing pixels removes pixels!

We’re discussing a website here, so there’s no physical shrinkage to consider.

On 4/8/2024 at 1:14 PM, John Rostron said:

If you are using these images in a website,  then the dpi is irrelevant.

 

[thomaso]

16 hours ago, nickbatz said:

Everything is closer together = pixels are less spread out = everything is sharper = the earth is round.

29 minutes ago, nickbatz said:

ridiculous argument

small = sharp?  ... Isn't sharpness a question of contrast ... rather than size?

[David in Яuislip]

1 hour ago, thomaso said:

Isn't sharpness a question of contrast

Yes, edge sharpness, what we used to call acutance in the darkroom days

Lots of info about sharpening on the internet, much of it twaddle
If you read an expert, the late Bruce Fraser, he recommended using Luminosity blend mode and restricting sharpening to midtones between levels 25 & 200 which translates to 10 & 78 in Photo Blend Ranges. The attached macro attempts to do that, adjust USM parameters as required

As an example I took an image 5076x3806 px and resized using Bicubic to 800x600, saved as png without sharpening, with Photo USM at radius=0.3, factor=2 and the macro. These were converted to webp at quality 65 using cwebp

File sizes:
webp65    png    
34386    640974    SharpenNone
55290    788657    SharpenUSM
53586    761879    Sharpen25-200

Now a visual example. Note how a browser will follow windows display settings and upscale images. This can be corrected with Javascript. Note also that sharpening can be done on an unsharpened image by the browser. It's not much different from using an image editor and there is a good saving on filesize. File zipped  to avoid forum mangling
Bonus point if you know the location

Sharpen25-200Lum.afmacro SharpenDemo.zip

[nickbatz]

There's plenty to say about sharpness, sharpening, contrast, resolution, destination formats... most of which are interactive. And scale is a hugely important artistic parameter, i.e. different pictures want to be different sizes (Claes Oldenburg demonstrated that effect pretty dramatically).

But here's the same picture shown large (a bunch of unidentifiable squares) and small (clearly [sic] a bush). No resolution change, just... well, I'm not sure how this could not be beyond obvious: if you're going to make a large print, you need to start with a sharp file (unless you're after a blurry effect).

[KarinC]

It is kind of interesting regarding pixels. I googled pixels, what size are they, etc. The answers were all over the place. Some say they are only a number and do not really exist, some answers say that they have a definite size, others say they don't. It is kind of curious what IS happening to them when you resize an image. Someone must know.

It reminds me of what my 94 year-old father said when he discovered google earth. "That is where the internet is, you know". "No, dad, I didn't know that", I responded.

[Alfred]

27 minutes ago, KarinC said:

It is kind of interesting regarding pixels. I googled pixels, what size are they, etc. The answers were all over the place. Some say they are only a number and do not really exist, some answers say that they have a definite size, others say they don't. It is kind of curious what IS happening to them when you resize an image. Someone must know.

The word ‘pixel’ is a portmanteau word from ‘picture’ and ‘element’. When represented in the physical world, the size of a pixel is going to depend on the characteristics of the output device: at any given display resolution, a pixel on a 28-inch screen is going to cover four times the area of a pixel on a 14-inch screen.

When you resize an image by changing the DPI but leaving the original pixels untouched, you’re simply specifying that the output image should be made larger or smaller by making the dots more loosely or more tightly spaced on the printed page. When you resample an image to maintain the same DPI for different physical sizes, you have to generate pixels for a larger image or discard pixels for a smaller image; the quality of the result will be heavily dependent on the resizing algorithm used.

[KarinC]

8 minutes ago, Alfred said:

The word ‘pixel’ is a portmanteau word from ‘picture’ and ‘element’.

Ha! I didn't know that! Thank you. 

The sources that say that a pixel is 1/96 of an inch would not be correct then - right - wrong?

 

Even the word portmanteau is a portmanteau. I love learning a new word.

[Alfred]

1 minute ago, KarinC said:

The sources that say that a pixel is 1/96 of an inch would not be correct then - right - wrong?

Back in the days before high-resolution displays, the Windows standard was 96 DPI and the OS X standard was 72 DPI (cf. 72 points per inch for type sizes).

[R C-R]

1 hour ago, KarinC said:

The sources that say that a pixel is 1/96 of an inch would not be correct then - right - wrong?

As explained here, the definition of pixel is greatly dependent on the context so, absent of any qualifiers that statement would likely be incorrect more often than not.

[thomaso]

19 hours ago, KarinC said:

The sources that say that a pixel is 1/96 of an inch would not be correct then - right - wrong?

Since 'pixel' is a word like cell, slice, piece, tile, etc. without an absolute size, the statement "a pixel is 1/96 of an inch" can be right or wrong. Since most screens nowadays have a higher resolution than 96 dpi the statement is rather wrong than right. A laptop, a smart phone, a smart watch and a digital billboard may show an image file of identical size (number of pixels) in quite different display sizes, regardless of a physical unit (inch, mm, feet, etc.). Accordingly they will show the same file with pixels of quite different sizes, tiny on the phone, huge on the billboard. The devices have different pixel density.

"a pixel is 1/96 of an inch"
compare: "a tile is 1/10 of a metre" … "a chess field is 2x2 inch" … "a piece is 1/12 of a cake" … "a slice is …

[Old Bruce]

19 hours ago, KarinC said:

It is kind of interesting regarding pixels. I googled pixels, what size are they, etc.

Pixels are squares* that hold one colour and only one colour. They can be any physical size. A 100 x 100 pixel image can be any size. Using that 100 x 100 pixel image make two paper prints. The first will be 1/4 of an inch square the second will be eight feet square, there will be 10,000 pixels in each paper print. The DPI/PPI will be different. In the first case it will be 400 DPI/PPI and in the second case it will be a teeny bit over 1 DPI/PPI**.

 

* Unless we are talking about video but we are not so they are squares.

** 1.041666666666667 according to my calculator. If the second paper print was 8 feet 4 inches square then the DPI/PPI would be exactly 1.