Reducing an images dimension produces fuzzy pixelated images

[cbolt]

Starting with a 4mb image (1200x1567px) and trying to crop and reduce image size. Pixel ratio to remain the same....fuzzy and pixelated everytime. grrr

I have tried all the suggested ways- resampling, bicubic and lanczos options. Why would my image be more fuzzy at a reduction in scale?

Edited January 11, 2019 by cbolt
to include image sample

[gdenby]

It appears to me that there are .jpg compression artifacts in both images. So by reducing the size, the sample blocks are merged into ever coarser averages. I don't do enough image processing to know, but I'll suggest  try de-noising the larger, or do a frequency separation before making the image smaller. 

[R C-R]

You reduced the image from 1200  x 1567 px to 306 x 400 px, so to begin with the aspect ratio is not quite the same, but more to the point you reduced the total number of pixels to around 1/15th of the original total.

Basically, that means all the color & sharpness info of each ~15x15 pixel area of the original has been reduced to about one pixel, so there is no way that pixel can retain all those colors or sharpness.

[JimmyJack]

This is what I get with Lanczos non-sep.
PNG (which is of course lossless) and JPG best (lossy nonetheless). But both are much crisper than the example above

 

[v_kyr]

As the preposters said or already indicated, this is also the reason why certain software optional offers to apply a little bit of auto-sharpening together when image resizing operations are performed.

[cbolt]

5 hours ago, R C-R said:

You reduced the image from 1200  x 1567 px to 306 x 400 px, so to begin with the aspect ratio is not quite the same, but more to the point you reduced the total number of pixels to around 1/15th of the original total.

Basically, that means all the color & sharpness info of each ~15x15 pixel area of the original has been reduced to about one pixel, so there is no way that pixel can retain all those colors or sharpness.

so how do I resize an image to retain its sharpness? Im accustomed to PS with the experience that reducing image dimensions didnt reduce its pixels but maybe it did and I was so accustomed to the process in PS that I didnt realize it? I'd love to make the shift from CC to AD/AP but am trying to keep my workflow tidy. Thank you for your message/help. cb

[cbolt]

4 hours ago, v_kyr said:

As the preposters said or already indicated, this is also the reason why certain software optional offers to apply a little bit of auto-sharpening together when image resizing operations are performed.

thanks for the message and feedback- does it matter at what point in process auto-sharpening is done? Do I do this to images that are resized for web use also?

best,cb

 

[cbolt]

10 hours ago, gdenby said:

It appears to me that there are .jpg compression artifacts in both images. So by reducing the size, the sample blocks are merged into ever coarser averages. I don't do enough image processing to know, but I'll suggest  try de-noising the larger, or do a frequency separation before making the image smaller. 

Thank you for the response and feedback- so should I not resample? Would this workflow be different of you were prepping images for web use? Is de-noising and frequency separation simply a work around to an app glitch? Resizing seems like a pretty simple/basic task to do on a photo editing app to have to do a work around to make it so. : ((

[cbolt]

4 hours ago, JimmyJack said:

This is what I get with Lanczos non-sep.
PNG (which is of course lossless) and JPG best (lossy nonetheless). But both are much crisper than the example above

 

thank you for your response. So it sounds like Lanczos non-sep and having PNG being my final image setting is a better route than jpg?

So save them as PNG files then reduce image using the Lanczos non-sep setting....is that what youre suggesting?  

I will have to ultimately turn these into web-friendly images but needed to start at the print quality resizing first.

Best, cb

 

[John Rostron]

I have to say that I cannot see any jpeg artefacts, but then my eyesight is definitely not what it was. I would agree though that jpeg artefacts could exacerbate the definition of your image on resizing.  Was the original image a jpeg? If not, then could you apply the reduction/resizing to the original non-jpeg image and see what that looks like?

John

[R C-R]

1 hour ago, cbolt said:

so how do I resize an image to retain its sharpness? Im accustomed to PS with the experience that reducing image dimensions didnt reduce its pixels but maybe it did and I was so accustomed to the process in PS that I didnt realize it?

You can reduce an image's dimensions without changing its pixels, but if you resample an image to reduce the number of pixels in it (its pixel resolution), sharpness will be degraded. See for example this Adobe Photoshop user guide page, in particular the Resampling section.

[v_kyr]

1 hour ago, cbolt said:

thanks for the message and feedback- does it matter at what point in process auto-sharpening is done? Do I do this to images that are resized for web use also?

You would do that as the next step after a size resample (decrease or enlarge an image). See for example here ...

... yes for web images too if a resampling descreases their visual sharpness here.

[JimmyJack]

2 hours ago, cbolt said:

thank you for your response. So it sounds like Lanczos non-sep and having PNG being my final image setting is a better route than jpg?

So save them as PNG files then reduce image using the Lanczos non-sep setting....is that what youre suggesting?  

I will have to ultimately turn these into web-friendly images but needed to start at the print quality resizing first.

Best, cb

 

As @R C-R said if you're resampling then there will be a certain loss of quality. This is why we have different resampling methods to choose from. Some do "better" in certain situation than others. In this case, to my eye, Lanczos non-sep is the winner whether you go with PNG or JPG. The difference there is that the JPG will further compress your image when saving.... even at the 100% setting (it looks a tiny bit softer than the PNG). PNG does not.
So, if saving a couple kb is not a huge concern, yes I would go with Lanc non-sep png. 
I did it in one step on the export.
(if you want to add additional sharpening (i.e. Unsharp Mask Filter) along the way (or after) that's up to you . Try it, see what happens.)

Hopefully the "print quality resizing first", is more of a resize without resampling, so not fraught with danger ....(at a safe to assume 300dpi, you'd be at ~4"x5")

[cbolt]

Thank you all- I have been using PS for years and really want to love AD/AP but this is proving to be a much more complicated transition than I expected. I understand there is a learning curve here but this is frustrating. I have to say at this point the workflow for resizing/prepping for the web in PS works flawlessly and I can complete it in about 3 very straightforward steps in about 30sec per image. I am convinced there is some aspect of this process that I am missing- maybe that's the learning curve of a new app?  I'll look forward to more posts about this as I'm sure I'm not the first to be struggling with this. best,cb

 

[R C-R]

20 minutes ago, cbolt said:

I have to say at this point the workflow for resizing/prepping for the web in PS works flawlessly and I can complete it in about 3 very straightforward steps in about 30sec per image.

It isn't completely flawless in PS either, but the reason it usually looks better may be that Adobe is using some kind of undocumented adaptive pre-sharpening & other adaptive methods to optimize 'Save for Web' exports. Several years ago, I read a blog post about this possibility. Unfortunately, I can't remember all the details, nor (of course!) can I find that post now, but the gist of it was that this was done based on the change in the pixel resolution, & is not quite the same thing you get when choosing a resampling algorithm or file type.

I am not sure how true this is, though.

[gdenby]

14 hours ago, John Rostron said:

I have to say that I cannot see any jpeg artefacts, but then my eyesight is definitely not what it was. I would agree though that jpeg artefacts could exacerbate the definition of your image on resizing.  Was the original image a jpeg? If not, then could you apply the reduction/resizing to the original non-jpeg image and see what that looks like?

John

If I use the Mac's default image viewer, Preview, and zoom in 5 times on the larger image, the artifacts are very evident in the sweater. Also a number around the mustache. The smaller one also has some color bleeding from the routine trying to average color changes around the left ear.

Also, It seems like there may be some banding in the larger, as if it was scanned.

[John Rostron]

1 hour ago, gdenby said:

If I use the Mac's default image viewer, Preview, and zoom in 5 times on the larger image, the artifacts are very evident in the sweater. Also a number around the mustache. The smaller one also has some color bleeding from the routine trying to average color changes around the left ear.

Also, It seems like there may be some banding in the larger, as if it was scanned.

If I look at enlarged views of these images, I can indeed see the artefacts where the sweater meets the background. I was only looking at the 100% view. I feel that if you can't see a problem at 100%, then it is not a problem.

John

[gdenby]

13 hours ago, R C-R said:

It isn't completely flawless in PS either, but the reason it usually looks better may be that Adobe is using some kind of undocumented adaptive pre-sharpening & other adaptive methods to optimize 'Save for Web' exports. Several years ago, I read a blog post about this possibility. Unfortunately, I can't remember all the details, nor (of course!) can I find that post now, but the gist of it was that this was done based on the change in the pixel resolution, & is not quite the same thing you get when choosing a resampling algorithm or file type.

I am not sure how true this is, though.

Perhaps:

"From what I've seen, there is no "general" method for calculating the quantization matrix. The quantization matrices are created in a way where after much experimentation, those numbers are the best that give a good SNR and gives good perceptual quality leveraging a decreased file size. Depending on who you talk to, or what product you're using, some use their own quantization matrices. For example, Adobe uses their own quantization matrices in products such as Photoshop and the generation algorithm is a trade secret - we don't know how they derived their quantization matrices."

from: https://stackoverflow.com/questions/29215879/how-can-i-generalize-the-quantization-matrix-in-jpeg-compression

[R C-R]

15 minutes ago, gdenby said:

Perhaps ...

Thanks for the link. It isn't the blog post I read, although the "trade secrets" & "quantization matrix" parts sound right. IIRC, the author compared PS 'Save for Web' export files to those created using other methods PS offers for exports, sharpening, & so on. He concluded that the 'Save for Web' algorithm(s) did something tricky that could not be done with manual adjustments, & it varied depending on the change in resolution & (maybe) the content.

That sort of jibes with some other stuff I read about changing the resolution before using the 'Save for Web' feature to get the best tradeoff between file size & quality, but I never was able to find enough info to decide if he was right.