Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Best way to resize images?
#1
Looking for the best way to resize an avatar for a website that won't result in any loss of quality (if possible). 

So far from playing around I've found 3 ways to do this.

a.) Drag and drop the image into GIMP > Image > Scale Image > Set Width & Height. (This makes the image look blurry even though I'm only reducing it by 50 pixels wide and 100 pixels tall. 

b.) Drag and drop the image into GIMP > Image > Canvas Size > Set Width & Height > Use the move tool in the canvas size window to move the image around until I've found the right spot > Click Resize. This looks much better than a.)

c.) New > Set Width & Height > Drag & Drop Image > Use Move Tool To Reposition Image. Looks clearer than a.) as well. 

So is option b.) and c.) the best way to do this absent loosing any image detail? 

Thanks.

Maybe some kinda sharpen filter will make images look cleaner / sharper afterward?
Reply
#2
With B and C you don't change the size of the image itself, you are just copping it or adding a margin. If you are cropping, there there is a 3rd option, the Crop tool.

Downscaling can generate a bit of blurring, there are several sharpening filters in Filters>Enhance that can mitigate this.
Reply
#3
When using Image > Scale image, choose either Nohalo or Lohalo for interpolation. They give less blurry results than Linear or Cubic.
Reply
#4
(04-19-2019, 07:03 AM)tmanni Wrote: When using Image > Scale image, choose either Nohalo or Lohalo for interpolation. They give less blurry results than Linear or Cubic.

I tried that. (Both of them). Can't say I noticed a difference.

(04-19-2019, 06:03 AM)Ofnuts Wrote: With B and C you don't change the size of the image itself, you are just copping it or adding a margin. If you are cropping, there there is a 3rd option, the Crop tool.

Downscaling can generate a bit of blurring, there are several sharpening filters in Filters>Enhance that can mitigate this.

Any idea what sharpening filter I would use? 

To give you an example of what I'm seeing when I use B/C options: https://i.imgur.com/iT7cfWc.png
If you take a look at the ear where the black arrow is pointing, you can see this sort've blurry pixel mess that surrounds the edge of the ear. Granted, for the purpose of demonstration, I used ctrl+mouse wheel and zoomed in at 250% to clearly show it, but It's still somewhat noticeable at 100%, just not as much. 

Strangely, I don't see this issue with option A -- it  just looks blurry another way lol.
Reply
#5
(04-21-2019, 08:49 AM)Fennec Wrote: To give you an example of what I'm seeing when I use B/C options: https://i.imgur.com/iT7cfWc.png
If you take a look at the ear where the black arrow is pointing, you can see this sort've blurry pixel mess that surrounds the edge of the ear.

Compression artefacts.

see: https://photo.stackexchange.com/question...about-them

Since the b) and c) option as described no scaling involved. My guess is those already exist and maybe with a new background become more obvious. This is an existing avatar ? and intention is scaling even smaller. Might be best to redraw completely.
Reply
#6
(04-21-2019, 08:49 AM)Fennec Wrote: To give you an example of what I'm seeing when I use B/C options: https://i.imgur.com/iT7cfWc.png
If you take a look at the ear where the black arrow is pointing, you can see this sort've blurry pixel mess that surrounds the edge of the ear. Granted, for the purpose of demonstration, I used ctrl+mouse wheel and zoomed in at 250% to clearly show it, but It's still somewhat noticeable at 100%, just not as much. 

Strangely, I don't see this issue with option A -- it  just looks blurry another way lol.

That blurry mess is JPEG compression artifacts. The tell-tale sign is the squareish pattern (JPEG compression first divides the images into 8x8 tiles):

   

So, your source image is crappy. With options B/C you keep the crap as it is. With option A, you blend the crap with the rest (because rescaling is computing new pixels values by cleverly averaging pixels in the image at the original size), so it goes under the radar (but participates in the general impression of blurriness).

As usual, Garbage In, Garbage out.
Reply
#7
(04-21-2019, 09:35 AM)Ofnuts Wrote: As usual, Garbage In, Garbage out.

How true! So, on scaling from .jpg, I always try first of all to reduce garbage--usually accomplished by blurring. Yes, that uses to improve results, at least blurring preserving edges, if possible, to get rid of compression artifacts. Those, if present, will get scaled and sharpened, and thus magnified as a rule.
I usually use plugin simply-bilateral, from the gimp repository, but no doubt there are alternatives...
After that, of course, I would scale and sharpen, in that order--or, for small images, where the default sharpen filter can be too strong (too thick black lines) instead:
- duplicate layer
- upper copy -> mode: soft light -> filter: high pass -> if needed, play further with mode and/or transparency.
Not perfect, but I hope it helps... Wink
Reply
#8
Blurring can improve results, but not for the reasons you think. It prevents aliasing by removing all the high-frequency detail that won't appear in the final downscaled image anyway.
Reply
#9
(04-21-2019, 09:35 AM)Ofnuts Wrote: That blurry mess is JPEG compression artifacts. The tell-tale sign is the squareish pattern (JPEG compression first divides the images into 8x8 tiles):



So, your source image is crappy. With options B/C you keep the crap as it is. With option A, you blend the crap with the rest (because rescaling is computing new pixels values by cleverly averaging pixels in the image at the original size), so it goes under the radar (but participates in the general impression of blurriness).  

As usual, Garbage In, Garbage out.

Ok, if I got a .PNG of this image, I assume it wouldn't have that jagged mess of compression artifacts? I think the reason I got a JPEG is because I needed the image to be a specific size and JPEG offers a smaller file size--at the expense of image quality I guess. It's strange, because when I zoom into the original JPEG, I don't see those artifacts. Dumb question, but I don't suppose you can convert a JPEG image to a PNG image since the source is already a crappy JPEG--it would just look the same?
Reply
#10
(04-21-2019, 10:01 PM)Fennec Wrote:
(04-21-2019, 09:35 AM)Ofnuts Wrote: That blurry mess is JPEG compression artifacts. The tell-tale sign is the squareish pattern (JPEG compression first divides the images into 8x8 tiles):



So, your source image is crappy. With options B/C you keep the crap as it is. With option A, you blend the crap with the rest (because rescaling is computing new pixels values by cleverly averaging pixels in the image at the original size), so it goes under the radar (but participates in the general impression of blurriness).  

As usual, Garbage In, Garbage out.

Ok, if I got a .PNG of this image, I assume it wouldn't have that jagged mess of compression artifacts? I think the reason I got a JPEG is because I needed the image to be a specific size and JPEG offers a smaller file size--at the expense of image quality I guess. It's strange, because when I zoom into the original JPEG, I don't see those artifacts. Dumb question, but I don't suppose you can convert a JPEG image to a PNG image since the source is already a crappy JPEG--it would just look the same?

Yes, once JPEG crap sets in it won't be removed by switching to a lossless format,it will just be carefully preserved. For CGI, PNG can do as good a job as JPEG with reasonable compression.
Reply


Forum Jump: