March 03, 2015, 10:16:27 AM

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Topics - helpful

Pages: [1]
EOS Bodies / Canon 5D3 vs. Nikon D600
« on: September 28, 2012, 09:47:03 PM »
I just received my Nikon D600 yesterday. I was expecting tremendous performance from another soon-to-be legendary Nikon/Sony sensor.

Well... I was quite disappointed.

With both cameras at ISO 12,800 and 1/1,600 exposure, with identical lenses (Sigma 85mm f/1.4 for Nikon and for Canon at f/2.8 ), what I get from the RAW data is shown in the attached images cropped at 100%.

This is really stupid, Nikon. I thought that with most of another year of tech gone by, and the legend of Nikon and Sony thrown in, that the D600 would perform better than the 5D Mark III. Boy, was I wrong.

The Canon image looks almost as smooth as butter under identical processing. The Nikon image looks like it came from a pocket camera or something like that.

I am also starting to get very suspicious about DxO Mark. Why do they have results for Nikon in a heartbeat all the time? Why do they measure dynamic range by trying to measure a theoretical definition (black point photo level to white point photon level) rather than trying to measure actual amount of detail in images that are under or over exposed by a certain number of stops?

I am thinking about doing a simple and truly mathematical measurement of image noise now, just to see if maybe I'm not giving the D600 enough credit.

The real thing that matters isn't DR or anything. Once the finite-dimensional subspace of data is fixed (which it is for raw files), the only thing that matters mathematically is signal to noise ratio, which is basically an aggregate of precision and accuracy, the two components of any recording technology, including photography.

Here's my plan, before I do it.

Take two or more successive exposures in RAW at the same settings, and repeat this process to obtain other pairs of image data with various under/normal/over exposure settings and various ISOs.

Then I will measure the actual noise by calculating the difference between identical images. The means of the data will be adjusted to account for a tiny variation in exposure times. Any difference between the images would be purely due to the random variation of noise.

Then I will use the old formula from science for relative error (observed-expected)/expected * 100 and then use the RE to calculate the signal to noise ratio.

And one number is meaningless. DxO Mark loves to give the highest ISO where (in their system of experimentation) the SNR falls below about 80-85%, a "critical point" of image quality.

But with cameras like the 1D X, the SNR barely falls any further for a long, long way.

However, cameras like the 5D Mark II fall off much, much faster after getting to this "critical point" even if their score reported by DxO Mark isn't very much different from the 1D X.

For example (made up numbers):

Camera A:
ISO 1,000 SNR = 90%
ISO 10,000 SNR = 75%

Camera B:
ISO 990 SNR = 90%
ISO 10,000 SNR = 45%

Obviously, camera A is 65% better than camera B, but the way DxO Mark reports things, there would only be a meaningless difference of about 1% of one stop in the "ISO score" of camera A vs. camera B.

So if I get time (tonight is a RARE few hours off for me), I will try to report the results. And I'll be totally unbiased. I use about half Nikon and half Canon equipment, and I have absolutely no grudges or favoritism on either side.


EOS Bodies - For Stills / Canon 1D X Raw Files -- Amazing
« on: August 21, 2012, 10:09:45 PM »
This image was an experiment I did back in July to see just how far the 1D X files could be pushed.

The photo was underexposed by at least 5 stops on a cloudy day and then pushed to the absolute maximum (+10.0 exposure) by Adobe Photoshop Lightroom 4.1. A little bit of noise adjustment was then made. But no curves, or any other fancy manipulation was done.

Enjoy the killer... the 1D X!!

Update: Same scenario, overexposed by 3 stops. Pulled to match the previous image.

The underexposed photo was at 1/1250th of a second.

The overexposed photo was at 1/5th of a second.

Let's count:

0 stops: 1/5th (pretend like this is ISO 100)
1/3 stop: 1/8th (iso 125)
2/3 stop: 1/16th (iso 160)
1 2/3 stop:: 1/32nd (iso 320)
2 2/3 stops: 1/64th (iso 640)
3 2/3 stops: 1/125th (iso 1250)
4 2/3 stops: 1/250th (iso 2500)
5 2/3 stops: 1/500th (iso 5000)
6 2/3 stops: 1/1000th (iso 10,000)
7 stops:  1/1250th (iso 12,800)

And the photo itself has several more stops of dynamic range in it.

Cool thing is the comparison with the ISO. Basically, this camera can take a single photo and capture an effective range of exposures as if multiple photos were taken ranging from ISO 100 to ISO 12,800.

There's nothing special about this image, but it has a good combination of rather dark shadow and much brighter highlights. My 1D X arrived on Monday, but just before sunset tonight was the first time I could take it out for a spin.

The photo was taken at ISO 12,800 and 1/200th. Most of the softness that is visible is due to a thin plane of focus or JPEG compression or slight motion blur, or all three. It's amazing for ISO 12,800. Update: the photo is a 100% center crop from the full 18 MP image file.

Another image: the "X" on my camera wants to say "hi"-- (this photo wasn't taken with the 1D X, obviously):

EOS Bodies - For Stills / Canon 1D X Arrived from Crutchfield
« on: July 09, 2012, 04:31:13 PM »
Anyone else receive one from someplace besides B&H or Adorama? Mine just arrived from Crutchfield today to my delight. Unfortunately, I won't be able to enjoy it for a while as I am out of town.

Delivered On: Monday,  07/09/2012 at 1:50 P.M.
Shipped/Billed On:07/03/2012
Weight:8.40 lbs

I just had regular ground shipping. It was sent prior to the 4th of July and arrived today in mid-U.S.A.

Since I am unable to do any tests until I return home, here is a very trustworthy analysis of the 1D X sensor by a well-known astrophotographer:

Astrophotography isn't my field, but it's very interesting.

Software & Accessories / Brightening/Correcting Underexposed Images
« on: June 06, 2012, 04:55:54 PM »
Canon community:

I would like your help with some mathematical/imaging research that I am doing. I am the developer for an unreleased algorithm for correcting underexposure from any type of image. I developed it on the side about nine years ago, when I was in the initial stages of some image compression research. The algorithm uses mappings from complex analysis (not "complex" meaning "complicated," but meaning the use of imaginary numbers) to create a seamless tone curve for any color, brightening or darkening it somewhat independently of how much information is available in the historgram. The RAW/JPEG bit depth doesn't matter, but the compression ratio is somewhat important.

The algorithm does NOT work miracles, and it is very "RAW" in that no post processing is done, so sometimes it produces hot pixels (that's why compression ratio is important, because anything that wasn't true data in the original file gets exaggerated when mapped using complex variables).

The algorithm does not differentiate between "shadow" or "highlight" brightening; it simply works by maintaining the exact same color hue, and adding color-independent brightness to it. The biggest benefit I believe is that colors are extremely pure and precisely maintained throughout the whole process. It's kind of like showing a slide--just brighten the projector to see very dim details and increase the dynamic range.

I am looking for some photos (small resolution, because that's not important; around 720px for the maximum dimension) to experiment with that aren't my own photos. I understand my own algorithm too well, and I feel like I can cook the results too easily if I use my own photos.

I am attaching an example of a photo that is two stops underexposed, and then the corrected version. (Note: the first image is actually at a greater than 100% magnification--I accidentally had my Matlab window too large when I took the screenshot.)

Thanks everyone.

The lower ends of ISO for the 5D3 are not necessarily spectacular improvements over low ISO from previous Canon cameras or versus the Nikon D800 (but very comparable). This is simply because low ISO performance has gotten as good as it can get already.

But the amazing thing about the 5D3 that doesn't show up in DxO-Mark style testing is how flat its image quality curve really is. Although it may reach SNR 85% at a somewhat lower value, its high image quality stays essentially the same for a range of several thousand ISO values.

Just to prove this, I am attaching an unedited ISO 3,200 100% crop from my 5D3. There has never been this little difference between ISO 3,200 and ISO 100 in any other camera. Period. Not even the D4.

As I have mentioned before, the 5D3 seems objectively better by three stops compared to the 7D, two stops better compared to the 5D2. At low ISO there is no point arguing, but the difference is obvious anywhere above ISO 800.

Canon has a winner on their hands. In my mind, the values of the 7D, 5D2, and all previous cameras have just dropped by about 90%. If you are trying to decide between a 7D, 5D2, or the 5D3, there is absolutely no comparison. Three stops means eight times better (and to me, worth eight times as much) as the 7D. Two stops means four times better (and to me, literally worth four times as much) as the 5D2.

And the improved autofocus and all that is the incredibly delicious icing on the cake. For those who don't rely heavily on autofocus systems, there is still reason enough to buy the 5D3. And for those whose jobs depend on taking rapid-fire sequences of randomly moving targets in perfect focus, the improved autofocus added to the image quality makes the 5D3 the perfect camera. That was something the 7D could do to some extent, and which the 5D2 could not do at all.

Pages: [1]