8MP vs 10MP, an informal test

Hi all,

I've just placed on DIMi the results of an informal test of upsampling
an 8MP Canon 350D image to the same resolution as the 10MP (roughly)
Nikon D200. The article is here:
<http://www.dimagemaker.com/article.php?articleID=460>

Comments very welcome on whether this is meaningful, a load of rubbish
or whatever

BTW a reminder to get your first entries into the DIMi panorama
competition for February.
<http://www.dimagemaker.com/comps/realviz1stq2006/panorama.php>

Cheers,

Wayne

Wayne J. Cosshall
Publisher, The Digital ImageMaker, http://www.dimagemaker.com/
Assistant Director, International Digital Art Award
Personal art site http://www.artinyourface.com/

 

The D200 image is cleaner, while the upsampled 350D image contains a lot
of artefacts (even the not upsampled 350D image contains many
artefacts). Not sure if this is a JPEG compression problem or some 350D
problem though.

 

There are some serious JPEG artifacts on all those images. That
removes all meaning from an already mostly useless comparison.

 

SNIP

Noise reduction?

Bart

 

It's not noise, more something like a mix of aliasing, compression
artifacts, oversharpening.

 

You'll be very hard pressed indeed to realize any significant
improvement with a 10/8 increase in the total number of pixels. The
increase in the number of pixels in either direction is only a very
modest square root of (10/8) =1.118 . So, in an area where the maximum
number of either horizontal or vertical lines was limited to say 100,
then this would increase to only about 112 ...

 

On the other hand the print area is 25% bigger - significant enough.

 

Hi all,

I'm going to redo the test. I'm sourcing two identical lenses in the
respective mounts and will redo the whole thing. I will also put tiff
crops up on the site so for those who wish to compare the actual pixels
with no jpeg compression to put on the web page, they can do so.

Thanks for all the observations.

Cheers,

Wayne

Wayne J. Cosshall
Publisher, The Digital ImageMaker, http://www.dimagemaker.com/
Assistant Director, International Digital Art Award
Writer and educator in graphic design, photography, digital technology
Personal art site http://www.artinyourface.com/

 

The old AF bar chart resolution charts had the patterns varying by about
11%. This was done because it was believed the minimum perceived change
in resolution was about this value. So a 12% increase in resolution
should be right about at that minimum perceivable change- you probably
wouldn't see it.

 

Unless you're ptintout is one pixel tall, it won't seem like 25% bigger.
More like 12% bigger.

 

Did you read what I wrote ? The print area will be 25% bigger.