On Fri, 11 May 2007 08:57:35 -0700, "Dave Johnson [MSFT]"
<davejoh@online.microsoft.com> wrote:
>I think it might be a problem with your video driver (are you using an
>Nvidia graphics card?). I don't see that problem on my PC using your sample
>image, and I've never experienced zooming quality issues with Photo Gallery.
I do have a Nvidia card, installed the latest drivers and the quality
on zoom in Photo Gallery is about the same as any graphic viewer.
One BIG problem with zooming in on any raster or bitmaped based image
is the more your zoom the more pixelated the image becomes. This of
course is due the image itself since it is comprised totally of
rectangular pixels, thus the more you zoom the more staircase effect
you'll get.
For more:
http://en.wikipedia.org/wiki/Raster_graphics
While it has nothing to do directly with Photo Gallery or any
application that zooms in on the fly, one trick you can apply if you
must enlarge a image much and intend to keep the enlarged version is
to enlarge it in small steps. In other words if if want to make a
image 50% larger, rather then making it 150% in one step instead make
it 110% larger in the first pass then enlarge that file and so on
until you get the degree of enlargement you need. BEFORE attempting
you should save in a lossless format such as .tiff so not to further
reduce the quality due to multiple compression steps. Only once you
are satisfied with the final result should the last file be saved in a
compressed format such as .jpg.
WHICH program you use of course makes all the difference. For example
Photoshop is very good at this, lessor applications are not. What
method you use also impacts the overall quality.
The "secret" is how the image gets interpolated.
While technical and mostly fairly advanced math, the "how" is
interesting reading: Also explains why they teach Algebra in high
school. :-)
http://en.wikipedia.org/wiki/Bilinear_interpolation http://en.wikipedia.org/wiki/Bicubic_interpolation
The second method (if available in your graphic application) is more
powerful, but slower, however if often gives superior results. This
applies to both still images and video color correction and
transcoding.
The reason is for raster based graphics when you enlarge them it is
better to sample not only ajoining pixels, ie what a pixel's neighbor
is above, below and left and right, then caculate on this, as it is
done in bilinear interpolation, but also to sample what pixels are on
the diagonal (Bicubic method) thus caculating the square or using a
two dimensional caculation which often yields better results picking
up more predominating tones since under this method some pixels get a
weighted average of this caculation. How "good" any enlargement looks
is not only impacted by how much it is enlarged, but how the added
pixels get caculated. If you do it in small steps, the guess work is
less and you're less likely to introduce error. Now you know. ;-)