Sorry to dredge this up again, but I think I may have discovered the reason why some people might have been disappointed by the quality of the M8 crop from my test. I did not explicitly state that the crop posted was effectively at 200%, not 100%. At the time I thought this was clear from the description of the test method, but perhaps it wasn't clear enough. I have inserted text into my comment on page https://www.fredmiranda.com/forum/topic/470256/143 to make this clearer. It's 200%, and contrast between adjacent pixels is very high - high enough to guarantee that the camera was capturing detail at full capacity. Apologies for any confusion.
It is still not in focus. The M8 would be giving moire or colour artifacts with black letters on a white label. With the M8, what you focus on is in focus and the depth of field doesn't seem to bring close to in focus into focus.
For example, here is a file you have seen. I focused near Guy or the grave Nicolini. Even at F8 and this distance, the areas behind this marker are not as sharp. The depth of Field is not going to make it sharp like a properly focused image. Also notice some of the colour artifacting in the letters. None of your test shots ever got sharp enough to do this (colour aliasing/moire). I suspect it was just focus errors or camera shake that prevented the shots getting ctitically sharp.
http://www.robsteve.com/FM/L1040403.jpg
A full rez, 100% crop
http://www.robsteve.com/FM/L1040403-2.jpg
A 200% crop.
http://www.robsteve.com/FM/L1040403-3.jpg
BTW, with a Leica CCD image from the DMR or M8, you can tell at 100% if it is in focus or not. It it looks like it has a bit of an anti-alias filter thing happening, it is just out of focus.
Please note: what follows is an attempt to understand a question about the M8's focus and resolution in my crap flawed tests, which I find interesting. I hope some others find it interesting too. However, if you don't want to read about the M8's resolution and the discussion of whether my test could have been better focussed, LOOK AWAY NOW as I am not trying to start another row.
Here goes: in earlier posts I mentioned that I thought high contrast between adjacent pixels across a significant area of the image suggested that the area was in focus. I stand by that. The red dot on the sticker that I focussed in my test is a good example of this idea. Here are some crops:
I would still like someone more experienced than I to explain how an unfocussed lens can provide transitions from one colour to another like these. A one pixel transition is surely the norm at perfect focus, while there may be places where the edge is in line with the pixel boundary and the transition pixel is closer in value to one or other side of the line, giving something approaching the theoretical maximum resolution. In the second crop above we see an example of that theoretical maximum being approached. Failure to focus accurately does not allow specific examples of interpixel contrast approaching the theoretical maximum.
Another way of looking at this is to insert text using the text tool in Photoshop to see where the theoretical maximum resolution for this text lies. There are many variables, but using the metrics of the recorded text, specifically, height, x-height, tracking, kerning, density and so on, we can get a rough idea what the text could look like:
I think this illustrates that we shouldn't expect too much of the M8's 10 megapixels in terms of what they can resolve. 10 is 10, not 16 or 22. This crop shows that there may be a little moiré in the middle of the text. Remember that this is under tungsten light inside a display cabinet. A sharper subject material might be unkinder to the lower resolving camera.
I also think that it raises the question raised originally by the brolly crop in my first crap flawed test. The spokes appeared to have been carefully removed:
Is the M8 applying a 'dust & scratches' type of noise reduction to JPEGs in order to compensate for moiré? If so, can it be switched off without using raw?
Edited by brainiac on Dec 24, 2006 at 01:11 PM GMT
Rob - I don't see any inconsistency between your results and mine. The circumstances of the pictures are pretty different and, as far as I know, you didn't repeat the shot on the 5D so it's not really capable of comparison. Don't forget, my test was unfortunately done in indoor light with a longish exposure, so noise and noise reduction could be a significant factor.
I would be very interested if anyone can try to demonstrate whether any in-camera noise reduction hurts fine detail in M8 JPEGs. Comparing JPEG to raw in both good and bad light might be quite a revelation, and could be another reason why my test is regarded as unrepresentative by some. How about that: only two shots required in the hunt for in-camera NR.
Just post a link to the full jpegs. I think most of us have the experience to tell if a file is in focus or not, whether it is shot with a 2mp camera or a 22mp camera. I am not arguing resolution with you, just that your shots have been out of focus.
In regards to resolution, more pixels does not make a better picture. You need good pixels and a sharp, well focused lens to begin with. Here is a shot from a 2mp Kodak sensored EOS D2000, with a Leica lens, shot long before the alternative forum existed. You don't need to go to a 400% pixel view to see this is in focus or not.
robsteve wrote:
Dirt, shadows? I am not sure what black spots you are referring to.
In the shot of the cemetry with another gentleman (Guy?) in, bottom half of the photo are brown leaves. I see pepper powders - black single pixels, on every leaf.
Pondria wrote:
In the shot of the cemetry with another gentleman (Guy?) in, bottom half of the photo are brown leaves. I see pepper powders - black single pixels, on every leaf.
Bugs eating them? Did you look at the full size DNG or are you referring to the jpg posted in the previous page? The leaves are quite small and you just may be seeing the air space or ovelap between leaves.
I totally agree. But a camera with more pixels is likely, where the lens and filtration are not constraining factors, to offer higher resolution.
You do need to view at 100% or higher to see whether and where focus/lens resolution reaches the limit imposed by the file resolution.
I looked back at your original jpegs. It looks like the M8 was focused on the "store" decal on the window.
There is also a discrepency between your jpeg sizes and what I got out of the M8 in Yosemite. On the first few images I had the camera on fine jpeg and the file sizes were rangeing from 4.6 to 6.8 mb. your jpegs seem to be only 2.5mb. Your M8 may have not been shooting in high quality mode. If you had just shot DNG, there would be less room for error in camera setuo, such as jpeg quality, sharpness and other things that affect the image quality.
I suggest the next time you do a test, set the camera on large fine jpeg, plus DNG.
robsteve wrote:
On the first few images I had the camera on fine jpeg and the file sizes were rangeing from 4.6 to 6.8 mb. your jpegs seem to be only 2.5mb.
Robert, a landscape shot, raw or compressed needs normaly more space than an interior shot. The leaves on the ground of your Yosemite contains more detail than the black window in Brainiac shots and I would look there for the difference.
Everybody can test this at home with their own shots at home. Go make one shot in a forest and another with a large blue sky or a brick wall, compress them using jpg 8 or 10 and see what file size you will get on your HD.
The m8 is smaller and lighter and I think a little bit less resolution wouldnt hurt. Maybe somebody could make another test between the two cameras and see if the m8 is really much better than on Richards test?
bathman wrote:
Robert, a landscape shot, raw or compressed needs normaly more space than an interior shot. The leaves on the ground of your Yosemite contains more detail than the black window in Brainiac shots and I would look there for the difference.
I am not disputing that the 5D has more resolution, just that Richard's images have not been representative of the quality I have seen from my own shots with a M8 and other files I have seen from a M8. I realize the jpeg compression will change file sizes, but shooting DNG would bring that variable out of the equation.
It would also have helped if Richard shot his test with a Leica lens rather than a third party lens. The Leca 21mm has no peer in the third party M lenses. A more fair test might have been pairing a good 50mm on the 5D, against a 35mm Summicron ASPH. A 21mm lens will never resolve like a much 28mm lens. The difference between the 50mm and the 35mm will be a bit less.
I would do these tests myself, but I do not have a 5D, just a 1D to compare. I actually don't know anybody with a 5D around here. I am also still waiting on my M8 to arrive.
I think we went through all this in the DMR Bible, except we were comparing the DMR and the EOS 1D MkII. The conclusion was the EOS had more resolution, but the overall image quality of the DMR was better. I think we will find the same in regards to the 5D/M8 if sombody familiar with shooting both does some comparisons. I think Carsten promised to do this.
