PDA

View Full Version : resolution tests vs mtf tests



smelni
10-31-2009, 12:12 PM
I first want to thank Barry G for his comprehensive rez test and alias article for the 7d - really informative

but i also want to throw a little curve ball here - very often resolution is not nearly as important to image quality as MTF. MTF measures the reproduction of different frequencies (number of lines per inch) throughout a video system.

Since the rez test on the 7d shows a dramatic falloff into aliasing at a certain (and fairly low) resolution, that doesnt mean that the resolution that it is reproducing isnt amazing (or that it is amazing for that matter)

think of it this way - if you are trying to reproduce many lines per inch, the 7d aliased and therefore will have a poor if not non existant mtf at that frequency. But not take a medium frequency (medium number of lines per inch) and how does it perform. This is a real important question and speaks to the actual curve of the mtf for this camera system. The optical low pass filter is usually tuned to the curve on the mtf to cut off information when the mtf is so low that is should be reproduced as grey instead of aliasing occuring. We know that the 7d doesnt have a well placed OLPF for video. But how does the mtf curve behave before that point.

WHY in the world am I bringing this up. Well, lots of people are asking why the images looks so damn good, and everyone loves the images, and compliments all around - yet the camera has shown to have less the HD rez. The mtf MIGHT (I dont know) explain this. Perhaps the curve is so well shaped before frequencies at which aliasing occur that the image looks amazing EXCEPT For high frequency detail. PERHAPS the mid and low level detail are being so accurately rendered that it looks amazing.

I am going to try to run some mtf tests on the camera but I cant promise when - but if anyone has the capability of measuring the 7d mtf vs a "real" HD camera that would be great. I predict that we might see why the image from the camera is so pleasing even though rez tests arent having great results.

Tom Roper
11-01-2009, 08:36 AM
All you need is a slant edge target and Imatest image testing software. Unfortunately, not much is enlightened rather just supports what the resolution charts have been saying. The 5DMkII MTF50 comes out with slightly higher MTF numbers, 620 horizontal, 756 vertical than what Barry reports, but the other video cams report significantly higher as well.

For comparison MTF, an XH-A1 is about 800 lines horizontal, 700 vertical, an HV10 roughly about the same as the XH-A1, and an EX1 930 horizontal, 1100 vertical.

Whatever makes the images of the 5DMkII so well, I find them excessively aliased, and not so amazing, and thus I don't agree with the premise they are so pleasing to everyone. I wish they were. I'd like nothing more than to unload the EX1 and use my 5DMkII for everything. MTF 50 tests (which I have performed) don't shed any particular graces on the DSLR.

Edit:

Actually some of the 5DMkII videos are amazing, but to me it owes to saturation and contrast levels which are high, and of course the shallow depth of field. Resolution is not everything.

smelni
11-01-2009, 09:25 AM
my point would be to look at the shape of the MTF curve rather then the limit of the MTF to see if something "pleasing" is happening in the lower end of the curve

Tom Roper
11-01-2009, 10:20 PM
I understood your point. I have these MTF curves saved, and will post them later this week. Unfortunately, Monday is my infusion day at the hospital, radiation in the afternoon. 1 more week to go.

smelni
11-02-2009, 04:40 AM
well sounds like you have more important things on your mind - i wish you the best of luck - its great that you only have 1 more week. seriously man, best wishes