How we test resolving power

iso12233.jpgLike all photography review sites we too use the standard ISO 12233 test chart to evaluate resolving power. Our more recent improvement has been to use the new modified ISO 12233 chart by applied image, which includes up to 4000 lph (Lines Per Picture Height) for 10MP+ cameras.

To analyze the results, we use the Imatest program which outputs numeric results. We prefer Imatest as opposed to the standard practuce of looking at the point where the lines get dense, as we find this approach too subjective. Imatest is a more objective tool for analyzing resolving power, and it can also be converted to many other measurement standards (MTF, LW/PH, etc…) as indicated in the following table of some of our results:

MTF50
Camera
LW/PH Vertical
LW/PH Horizontal
lp/mm Vertical
lp/mm Horizontal
Nyquist MTF*
Nikon D200
1831
1856
57.9
58.7
0.264
Canon EOS 5D
1992
1996
41.6
41.7
0.175
MTF @ 30 lp/mm
Nikon D200
0.732
This MTF Data could be compared with some MTF lens plots

(Taken of Vertical)

Canon EOS 5D
0.946

The short glossary below gives some information to help understand more about our results table.

MTF: Modulation Transfer Function characterizes response of a device to a shorter frequency. A 100% or 1 MTF response would mean that the device has resolved 100% of the given frequency. A 50% or 0.5 MTF response would mean that the device has only resolved half of the given frequency. In digital cameras however, resolving power is displayed as line width per picture height (LP/PH) with a 50% MTF response.

MTF 50%: This is the chosen response by which we measure how much LP/PH a camera can resolve. MTF 50% was chosen because MTF 50% is the point where cameras start to decrease resolving power. Imatest website defines MTF 50% as “ an excellent measure of perceived image sharpness because detail is diminished but still visible”.

LP/PH: line widths per picture height is the spatial frequency unit we use to describe a camera. LP/PH is an excellent camera measurement unit because it takes into account the picture height, as long as the ISO 12233 chart fills the frame during capture. Unlike an APS sized sensor, a full frame camera (with the same lens) would have to be closer to the chart, as in the real world. Therefore the full frame camera has an advantage in the LP/PH results.

LP/PH can be converted in to LP/MM with the following equation:

LW/PH / sensor height in mm / 2 = lp/mm

lp/mm: This is the most common spatial frequency unit; it simply describes how many line pairs there are per mm. We use this unit when we want to describe the camera resolving power in mm without considering sensor size. With this unit a camera with a small sensor and higher pixel density can produce better results than a full frame or large sensor camera with smaller pixel density. This unit is considered to be less faithful to real world results because it only takes pixel density into account.

MTF@30 lp/mm: Most lens manufactures publish their MTF lens reading for two lp/mm units, 30 lp/mm and 10 lp/mm. Although it is not very customary for digital cameras, we chose to show MTF response a camera makes in 30 lp/mm. You can compare this number with some lens manufacturers MTF plots.

Nyquist MTF: Nyquist frequency is the highest point at which a camera can capture any useful signal. Any data captured above this point is simply unusable. This unusable data is eventually transformed in the final image to what we call – color aliasing or moiré patterns or false colors. Nyquist MTF is the term we use (and so does Imatest) to describe the camera’s response at Nyquist frequency. A high number (between 0 to 1) will indicate that the camera might exhibit high levels of false colors. A low number will indicate the camera might not suffer from noticeable false colors (Imatest indicates that aliasing effects may become serious over 0.3).

But, there is a catch. Camera makers are concealing false colors and moiré patterns by using special filters that cover aliasing problems with gray color and by reducing chroma level at those areas. This is not something this particular test can resolve. We use our demosaicing and false color test to evaluate this particular filter.

Anti Aliasing Filter: Because of high frequency signals that are unusable above the Nyquist frequency of the sensor, camera makers are using special optical filters that are placed in front of the sensor (a.k.a low pass filters). The purpose of these filters is to reduce the frequencies that reach the sensor into the frequencies range the sensor can sample. Thus, false colors could be prevented. However, the stronger the filter the softer the image will look, so camera makers are trying to find the balance between false colors, sharpening and image processing filters that reduce artifacts.