rjlittlefield wrote:mawyatt wrote:If these were just pure representations of the actual pixel level AD converter results [emphasis added]
I find it troubling to see that sequence of words used to describe a TIFF file. They suggest no appreciation for how much processing is done to generate
any TIFF file. At the very least, the sensor data has been demosaiced, gamma adjusted, and had its color profile translated from the sensor's native profile to some standard profile such as sRGB or Adobe RGB. In most cases, the image has also been sharpened and possibly noise-reduced. A typical TIFF image bears about as much resemblance to the original sensor data as a half-toned paper print does to an original piece of film.
None of this bears on the issue of file size, however. That must be a matter of compression. As far as I know, the main image in Nikon TIFF files is not compressed, so that part should be exactly the same from one file to another. But the TIFF file includes other parts whose lengths can vary. Given that much variation in file size (from 112.6 to 115.0MB), I'm thinking there may be something like a JPEG thumbnail embedded in the TIFF, with a compression ratio that varies depending on image complexity.
If you can DropBox a couple of files, I'd be interested to take a quick look.
--Rik
Rik,
"I find it troubling to see that sequence of words used to describe a TIFF file. They suggest no appreciation for how much processing is done to generate
any TIFF file"
Wrong suggestion, I have the utmost appreciation for the signal processing taking place. Point I was trying to make with this reference to the pixel ADC is your comment below.
"None of this bears on the issue of file size, however. That must be a matter of compression. As far as I know, the main image in Nikon TIFF files is not compressed, so that part should be exactly the same from one file to another."
I have seen small variations before in TIFF file size before but this was larger and caught my attention. My thinking was the file size was dictated by the pixel ADC levels, color/exposure info and some overhead, and thus should be somewhat constant for a given camera body.
"At the very least, the sensor data has been demosaiced, gamma adjusted, and had its color profile translated from the sensor's native profile to some standard profile such as sRGB or Adobe RGB. In most cases, the image has also been sharpened and possibly noise-reduced."
How would any of these actions effect file size if the TIFF was created without any compression?
"I'm thinking there may be something like a JPEG thumbnail embedded in the TIFF, with a compression ratio that varies depending on image"
This makes the most sense to me, it explains the small varying file size apparently varying with image details. However, I was unaware that a small JPEG or other image specific file was embedded within the TIFF format.
You should be able to run a simple experiment to see if this follows with your D800. I'll see if I can upload to Drop Box if I'm still active.
The files in question were of a silicon wafer at 1X, thus lots of very fine detail across the entire image field. I was stepping at 50 microns to get the everything in focus somewhere after stacking. So maybe the in and out of focus areas cause an apparent detail variation within an individual image? But this suggests that something "knows" these details within TIFF, thus affecting the file size by way of compression or embedded image!
Best,
Mike
)