Does resizing smaller effect clarity?

[jamezzz122]

I know that when making the resolution of an image larger, the clarity is worse. Is this the same when making an image smaller?

 

[phixt]

Short answer is no.

If you keep the dpi (density) the same, you won't lose quality unless you enlarge it again.
If you increase the dpi proportionally to how much you're shrinking the image, you'll retain quality even if you enlarge it back to the original size later on.
Reduce dpi, and you'll lose quality by shrinking.

 

[Hyperion]

The long answer is: Yes

Fig. 1

Resolution vs. Dot-pitch
Resolution is the number of sample points in a sample, this can be applied to any
data set, be it the linear samples of mono audio, or the planar samples of raster images.
Dot pitch is the relation between the sample-points and an SI unit of distance. Since
printing was invented before the metric system a very old metric called "Dots Per Inch" (or
dpi) has stuck even though some dot-pitches are expressed in metric (screen dot-pitch is
expressed as the size in mm between one dot and the next.

Resolution vs. Dot-pitch
From Fig. 1 it is pretty clear that even though the dot-pitch has not been kept constant
as phixt requested, that the amount of image information lessens and therefore adversely
effects image quality. Now I can immediately hear people saying that those are severe
resamples, that down-samples the original image by orders of magnitude. That is just to
illustrate the phenomenon more clearly. If one should resample a 1000px x 750px image
to 900px x 675px the quality loss would not be as severe as for an image of say
640px x 480px down-sampled to 576px x 432px.

But were're interpolating to a smaller set, why is this?
Yes, but even so, the sample points are discreet, not continuous, and therefore the step
effect of the pixels take their toll. Any digital photographer will tell you that a down-sample
goes hand in hand with an adaptive unsharp. Why? Because very fine textures are
typically diminished and small crisp speculars (i.e. glint in the owl's eye) become dull and
less pronounced.

So ... down-sampling even though not even remotely as dangerous as up-sampling is still
a topic to approach with care!

 

[Lethal]

Well, that's all very scientific sounding, Hyperion.

My experience in reducing pic sizes (e.g., making thumbnails) is that they come out much clearer if you step-down in several increments, adding a bit of unsharp masking each time.

 

[Hyperion]

Lethal: you are wholly correct in stating that you apply some sharpening algorithm when
reducing the size of a raster image. This however points to exactly what I have said before,
namely that, due to resolution loss fine detail diminishes (i.e. loss of image data).
However it is important to note that all sharpening algorithms attempt to restore contrast
to an image, and often overcompensate causing tell-tale 'sharpening halos'. Here's
the 'scientific' stuff:

Fig.1

Fig.1 illustrates the concept of sharpening, in the first case, no sharpening is neccesary,
and the 'sharpening curve' is therefore null, in the second curve there exists a linear
gradient this is corrected by applying the linear 'sharpening curve' shown, ditto for the
non-linear gradient.

Since statisticians thrive on a continuous frequency distribution known as the normal
distribution (also known as the Gaussian distribution after it's inventor), the Unsharp Mask
algorithm first applies a Gaussian transform to your image (obliterating all detail, after
which it then applies the appropriate 'sharpening curve' for an n-radius Gaussian gradient.

The effect is generally very convincing in that the image appears sharper, even though
the sharpness is an artificial contrast enhancement. The problem is that this contrast
enhencement cannot possibly work equally good for ALL pixels, and in many cases over-
sharpens leaving the halos mentioned earlier, Fig.2 shows these in some detail.

Fig.2

It is obvious that one should be VERY VERY careful with sharpening tools since they
inevitably alter the true image information, and subsequently add the risk of unwanted
distortions as shown above.

So let's say one has taken a digital photograph with a high quality camera (8MP sensor),
and after much editing decide to down-sample the image to 1200px x 900px @ 300dpi.
This should yield a print of 4" x 3" however reduce the file-size by down-sampling to
800px x 600px @ 200dpi, on screen the smaller image would look perfectly OK (if not
compared to the original which will have much more detail.

The surprise comes in print ... both images will print 4" x 3" however the 200dpi image will
look fuzzy and slightly out of focus compared to the 300dpi image. should the dot pitch on
the 200dpi file be increased to 300dpi the print dimensions will shrink to 2 1/6" x 2", this
will make the edges appear just as crisp, however there will be some details that will still
be lost due to the reduced resolution.

 

[pistola]

Couldn't you say perhaps... Shrink the size but increase the dpi to acomodate the image data to increase quality of the smaller image. But for his sake I think he means take say an image thats 4x6 scanned @ 300dpi and he resized the image say 50%. The "clarity" would be the same, although on the small image some of the image data will be thrown away, but to the eye... and the printer if printed the size it was resized too, will look as good as the larger 4x6.

edit--disregared that frist part, just tryed and that don't work...

 

[phixt]

Sorry for any misinformation I gave, I've learned a lot from this thread