line pairs per millimeter/pixels/sharpness

[chauncey]

What is the correlation between line pairs per millimeter and pixels in determining image sharpness?

In spite of the technology, the human eye can only perceive about 10 line pairs per millimeter
http://www.luminous-landscape.com/tu...harpness.shtml (http://www.luminous-landscape.com/tutorials/sharpness.shtml)

What would be the corresponding number in pixels...at what point do we visualize an image as being soft?

[JeffreyG]

It takes two pixels to form one line pair. So if you have a 4000x6000 pixel (24MP) camera you can resolve up to 2000 line pairs across the short axis.

If this were an APS-C sensor, then the resolution of the camera would be up to 2000 / 15mm = 133 lines per mm.

There are however two other devices at play before we even begin to need to think about what the human eye can perceive.
1) The lens. Few 35mm lenses can really resolve details smaller than ~80 lines/mm, so a sensor with a pixel density higher than 160 pixels/mm is probably not going to make a difference in the image.
2) The printer. Printers don't get much better as the pixel density goes past 240 / inch (This is that 10 lines/mm that the LL article was commenting on).

So let's go back and figure out the pixels needed to make a nice big 12x18 print from an APS-C sensor.

first, we want at least 240 pixels per inch on the printer. So a 12x18 print wants at least 2880x4320 (12.4MP).

But is the lens really resolving this level of detail? Probably not. Suppose you have a really cracking lens that resolves details down to 80 lines/mm with acceptable MTF. This is a mere 1200x1800 line pairs or 2400x3600 (8.6MP) pixels on the sensor. So in reality there is no gain in a APS-C sized sensor going past about 9 megapixels.

[RDKirk]

The author answered your question here:

It's worth noting that a 6 colour 1440 dpi inkjet printer (like the Epson 1270 / 1280 / 2000P Photo printers) when fed a 360 dpi output file, is capable of about 16 pixel per millimeter. This translates to 8 lp/mm

I would quibble, though, with the example of printer dpi resolution. Those super-high dpi claims refer to exactly what they say: Ink dots per milimenter, not pixels per millimeter.

An image pixel is an atomic element--the smallest element of the digital image--containing a single tone/color. Printer dots are actually "sub-atomic" particles--numerous overlapping ink dots laid on the paper to create an image pixel.

It may indeed take twenty printer dots at 1/1440 mm to build a single 1/8mm pixel.

[Wilt]

It takes two pixels to form one line pair. So if you have a 4000x6000 pixel (24MP) camera you can resolve up to 2000 line pairs across the short axis.

...But is the lens really resolving this level of detail? Probably not. Suppose you have a really cracking lens that resolves details down to 80 lines/mm with acceptable MTF. This is a mere 1200x1800 line pairs or 2400x3600 (8.6MP) pixels on the sensor. So in reality there is no gain in a APS-C sized sensor going past about 9 megapixels.


Thank you for reinforcing the lens' role in this whole thing. The average dSLR shooter has been so engross in pixel count that they ignore limitations imposed by lens optics!

[chauncey]

to expand a little on what Wilt was saying here's a pair from my tripod mounted 300 f/2.8 with/without a 2X TC
Screenshots from LR...magnification shown in upper left

http://i329.photobucket.com/albums/l383/chauncey43/sharp-600-1.jpg

http://i329.photobucket.com/albums/l383/chauncey43/sharp-300.jpg

Adding that TC sure screws things up...I'd rather crop down than add a 2 X TC