Erik's web

  • Increase font size
  • Default font size
  • Decrease font size

How many megapixels do we need?

E-mail Print PDF
User Rating: / 0
PoorBest 

Note: If you are reading this, you may check out this two articles, too:

 

I'm often beeing asked about how many Megapixels my camera has. The answer is 24.6, in short. In my mind Megapixels are not the most important issue, it may be more important how good those pixels are. My impression is that:

 

  • Different size prints need different number of Megapixels.
  • If we have more pixels than needed for a print size we get diminishing returns.

 

On the other hand, it is quite possible to make large prints from small files at least under certain conditions. The most important one is viewing distance. For instance I have A2 prints from my Konica Minolta 7D which are quite OK if viewed at say 80 cm distance, and also an 70-100 image hanging on my wall made from an 10 Mpixel image. The latter image is not really sharp, but looks decent at perhaps one meter distance.

 

In short I'd say, I would say that the following applies.

  • 6 MP is good for A3
  • 12 MP is good for A2
  • 24 MP is good for A1
  • 48 MP is good for A0

This may or may not be with a stretch. In practice we often can improve an image by cropping, and that's a goodbye to many fine pixels.

Human vision and Megapixels

Resolution of the eye

The resolution of the human eye is limited by the cones in the foveola. The foveola contain about 30000 tightly packed cones, which essentially correspond to pixels. The foveola has a diameter of about 0.4 mm, the focal length of the eye is about 22 mm. This correspond to about one degree of arc. The cones have a radius of about 2 microns, so the cones will be separated by about four microns in the foeva. We could say that sensor pitch of the eye is about four microns. This corresponds to about 0.6 minutes of arc, or 0.01 degree. For some reason the figure 1 minute of arc is normally used. At 25 cm this would correspond 0.073 mm

The Contrast Sensitivity Function of the eye

According to Imatest documentation by Norman Koren the CSF of the human eye is like below.

 

It seems that contras sensitivity of the eye has a maximum around 8 cycles/degree, which corresponds to a feature size 0.27 mm at 25 cm. To be able to separate two features we need at least something between them. If we call the features line we can say that the resolution of the eye is something like 7 lp/mm, but the eye is most sensitive for structures around 0.3 mm.

 

The 7 lp/mm means essentially that we need 140 pixels/cm. Would we use the separation of the cones which rather corresponds to 0.6 degrees of arc we would need about 220 pixels/cm.

 

Size 140 p/cm 220 p/cm
A4 8.5 MP 13 MP
A3 17 MP 26 MP
A2 34 MP 52 MP

 

Practical experience says that we need something like 180 ppi resolution for good prints. Elder Epson printers were optimized for printing at 360 LPI, which is close to the 140 p/cm figure. The DPI rating of printer like 1440 or 1280 DPI does only have to do with dot pitch but is not the same as the resolution of the printer.

Size 180 PPI 360 PPI
A4 3.16 MP 12.64MP
A3 6.32 MP 25.2 MP
A2 12.6 MP 50.4 MP

 

From the two tables above we may draw the conclusion that even if 12 MPixels is adequate for A2 the eye can actually resolve much finer detail.

 

Viewing distance plays a major role

If we double the viewing distance the feature size we can observe also doubles. We tend to see larger pictures at longer distances, so larger prints are less critical than smaller ones.

Edge contrast

I have made A3+ enlargments from cropped 6 MP pictures. They actually lack fine detail but are still perceived sharp. The reason for this is mainly that they are properly sharpened. The human eye interpretes edge contrast as sharpness.

 

Some examples

The images here are published under Creative Commons, Attribution, Non Commercial, Non Derivative Works license.

 

This cropped 6 MP images prints well on A3-plus, even if it actually lacks detail. It's a nice picture to hang on a wall but nothing to pixel peep at 25 cm distance.

Click on image for full size.

 

I printed this 6 MP picture in A2 and it looks good on my wall. It's clearly a stretch... If you look at details at short distance it's not crisp and contains some artificial detail.

Click on image for full size.

 

Some experiments

 

In reality the proof is in the print. I made two experiments recently.

Experiment 1

A photographer made a series of comparison shots between different equipments: here

I downloaded the Canon 5DII image and the Hasselblad H3II-50 image and printed both in A2 format using my standard procedure. The differences are summarised below:

 

  • Not a lot of difference
  • At short viewing distance (25 cm) the Hassy image has better detail
  • At longer distance (80 cm) Canon looks better to me, probably because of better DOF and more snap
  • It's quite obvious that digital processing plays a big role

I also asked a colleague who use to work at a professional lab, and he definitively preferred the Nikon image over the Hassy image. There were some suggestions on the LL-forums that these comparisons were poorly executed.

Experiment 2

I have some work in progress comparing scanned 67 Velvia with my Sony Alpha 900. I tried to reproduce the test target from

The Great Luminous Landscape 006 State-of-The-Art Shootout.

My image is  here

 

In my image I used a one dollar bill and a color checker in the same scale as in the LL shootout. I shot this on film and with the Sony Alpha 900. Still waiting for the film to return from lab... I tried a central crop from my image and printed in A4, the print corresponds to about 60x90 cm size for the full image. I made a similarly cropped print from one of the Hasselblad H1/P45 images (from LL shootout DVD). The results are quite different.

  • The "Hassy" image is significantly crisper when viewed at short or long distance.
  • The "Hassy" image has visibly more detail
  • Regarding tonality the images are close
  • The Sony image is quite sharp on it's own, but cannot compete with the "Hassy" image

No surprise, the test image is flat, DOF does not matter, tonality is similar. With all parameters constant the format with the more pixels wins.

 

An additonal test was done. The Luminous Landscape tests included shots taken at different apertures. The original images I compared were at f/8, but I made an additional print from the f/22 image.

  • The difference between the f/22 and the f/8 image was clearly significant.
  • The P45 image at f/22 seems to be in the same league as the f/8 image from  the Sony Alpha 900 at f/8.

The graph below may explain this partly. It was made with a Sony Alpha 100 (10 MP camera) and a Minolta Dimage 7D (6 MP) camera. Resolution for 50% MTF drops from around 2000 LW/PH to about 1200 LW/PH when stopping down from f/8 to f/22. This depends on laws of physics.

 

 

Last Updated on Sunday, 18 April 2010 20:44  

Statistics

OS : Linux u
PHP : 5.2.9
MySQL : 5.0.67
Time : 14:32
Caching : Enabled
GZIP : Disabled
Members : 1510
Content : 72
Web Links : 1
Content View Hits : 325555

Login

Banner1

J!Analytics