LaTeX graphics packages
LaTeX and its graphics packages do not touch the image data. TeX does not even provide
the reading of binary data. Thus LaTeX passes the image as file name reference to
the driver. Also most of the drivers are not image processing programs. They only move
the image data in a form appropriate for the output format. For example, dvips only
copies the PostScript file into the output PostScript file. Also pdfTeX often do not
need to unpack the image data and can copy the data to the PDF structures. Some PNG
files are uncompressed and compressed. But this process does not change the image data.
Very view drivers are able to resize an image, AFAIK Acrobat Distiller and
GhostScript does some obscure things with images like using JPEG compression for
PNG images. I do not know how this can be turned off. The related options given in Ps2pdf.htm
seems to have no effect.
But you are using pdflatex that does not change the image data.
This can be tested. Take a PNG image and convert it to PPM. Also embed
the image in a LaTeX document and run pdflatex. The program pdfimages (xpdf)
extract the image in PPM format.
convert image.png image.ppm
pdfimages test.pdf t
diff t-000.ppm image.ppm
image.ppm should be identical, if the PPM format
is the same (there is a binary and an ASCII variant).
** Different viewing programs**
However the programs
to view the images are different. The image viewer and the PDF viewer are usually
different programs that uses different methods for viewing. For example, a program
might use anti-aliasing, …
Scaling vs. resizing
There are no problems in scaling images in LaTeX independent of its method:
Only the place that the image uses on the page differs, the image data
remains the same.
A different term is the resizing of an image. Then the actual number of pixels
change. Of course, an image processing program cannot invent missing details
if the image is enlarged. It can only use better or worse methods to limit
the artefacts of resizing.
Screens have low resolutions comparing to printed media and are usually
stupid bitmap images of the pixels on the screen even if the original
data were high quality vector data (non-pixel fonts, vector drawings).
Some hints to get better quality:
- In some cases a vector screenshot program might be available, e.g. gtk-vector-screenshot.
- A special case are web pages. They could be converted to PDF by printing or
there are programs/sites that perform the conversion. But caution, PDF, especially
PDFs from screenshot programs might contain bitmap data instead.
- Higher screen resolution with larger font/symbol settings or using settings
for visually impaired people.
- A large monitor helps that allows large windows for the screenshot programs that
can only catch the pixels inside the screen.
- And it can make sense to turn off anti-aliasing and similar (ClearType) to
get clean pixel data. Thus that a black line is displayed by black pixels and
not by many gray levels at its edges. That makes it easier to optimize the
image for the final media.