Bit Depth, File Size, and Photo Quality

[marc_lieberman1]

When I scan a 35mm negative I get an enormous file- at least 67 megs.

At higher than normal bit depths, files can be twice that size.

An image file generated by a DSLR is a small fraction of that size.

However, the image quality, while arguably different, is not obviously

inferior. Why is that? What is the relationship between file size,

bit depth and photo quality? Is bit depth a scanning related concept

that is not relevant to digital capture?

[Craig_Cooper11664875449]

Marc,

 

Assuming your thinking about the Nikonscan 5000 and assuming it scans at 5000 DPI, which it doesnt (its a little shorter).

 

35mm frame is about 1" x 1.5" so 5000 x 7500 pixels = 37.5 Megapixels

 

Assuming your scanning in 16-bit mode, thats 48 bits per pixel or 6 Bytes per pixel, therefore...

 

37.5 MP x 6 B = 225 MBytes per file in RGB or 75 MBytes in Greyscale.

 

If you scanned in 8-bit mode, thats 24 bits per pixel or 3 Bytes per pixel, therefore...

 

37.4 MP x 3 B = 112 MBytes per file in RGB or 37 MBytes in Greyscale.

 

The actual numbers are probably about 90% of this as the scanner doesnt quite make a true 5000 DPI scan.

 

As for digi-cams its exactly the same concept except most top end ones appear to be around 12-bits per channel (36 in total) and if you look at camera reviews on a site like www.dpreview.com you'll see that the size of the sensor in pixels is much lower than what your scanner delivers.

 

Craig

[klaus.sailer]

Image size or megapixels is not a direct measuring factor for image quality. It's just the amount of data captured.

 

For the film scanning case, not all the captured information gives you more quality. The possible amount of real image information is (at best) limited by the information contained in the film. The film grain plays a big role in measuring the possible resulting real data from a film scan.

 

With digital camera catures, there are other limitations, like lens resolution etc. That's one of the reasons you shouldn't decide on a camera by looking at the megapixels alone.

 

Bit depth is also relevant for digital capture. It defines the steps between each of the possible brightness levels for each of the colors.

During processing, higher bit depth avoids losing information in between those brightness levels.

[byronlawrence]

... a side note for the math above.

 

there are 8 bits in a Byte, not 2, so a 16 bit depth image in color has 2bytes per channel (rgb) and hence the 6 total bytes per pixel. At 37.5 Mp is the calculated 225 MB (the end result is correct but the math in the middle had an error). half of that would be your 8bit per channel file - 112.5. Maybe there is some compression somewhere to get your 67 mb, or maybe you are scanning b/w 16bit which results around 75mb of info.

 

another difference is that most digital cameras shoot at a 12bit per channel depth. so you get smaller raw files. I know some scanners actually scan a a 12bit depth but I don't know if your scanner does or not.

 

anyway I thought I would add some more to the answers.

[Ed_Ingold]

Bit depth does not visibly improve image quality, but gives you more room to make adjustments. It makes sense to reduce the final file to 24 bits for faster printing and reduced storage requirements. I keep the 48-bit file as a digital "negative".

 

The LS-5000 scans at a nominal 4000 ppi. By various estimates, the true value is closer to 3700 ppi - which is still very good. The net resolution is limited by the density of the sensor array and the film itself.

 

The resolution of the scanner lens has not been published, but a single-purpose lens which operates at a fixed ratio can easly exceed 400 lppm, measured optically. Consequently, the lens has little impact on sharpness.

 

Film resolution is limited by grain, but more important by scattering (halation) of light within the emulsion at the time of exposure. The effect is roughly Gaussian, and has been estimated to limit resolution of low-contrast (i.e., normal) subjects to about 40 lppm.

 

Scattering does not occur within a digital sensor (though perhaps in the filters, and deliberately in anti-aliasing filtration). As a result, the visual appearance of sharpness (acuity, or edge contrast) is equivalent to a film scan with two or three times the pixel count.

 

The rest is history.

[byronlawrence]

yea to what Edward says. he makes important points.

[ellis_vener_photography]

a big difference, which no one seems yet to have mentioned here, is that the the image

out of the DSLR is a first generation image. Scan film and the digital image that results is a

second generation image. Sort of like taking a loupe to a negative to judge image quality

and then taking a loupe to a print of that negative and trying to compare using the same

standards.

 

By my reckoning there is roughly about a 3:1 ratio betwwen scanned and DSLR file size

image for equal iamge quality at 16 bits per channel (48 bits) --assuming the same lens

with at the same settings is used on both cameras.

[byronlawrence]

absolutely. in forensics I think they call them filters, or something of the sort. with each process data is lost, the lens, the medium, the manipulations, the print. etc. film has one more step to loose data compared to digital.

[sean de merchant httpw]

<p>There are many differences and many similarities between the two forms of capture. I find that one of the <a href="http://www.trenholm.org/hmmerk/PTT.html"> simpler and clearer explanations</a> of the visible differences was written by Harold M. Merklinger (see link).</p>

 

<p>The way he explains it is that digital retains contrast right up to its extinction resolution while film loses contrast and begins to be dominated by noise as one approaches its extinction resolution. Hence a digital image will be very clear in its fine details, whereas a 35 mm film image will not be as clear in those same fine details. The film will capture finer details, but with a lower signal to noise ratio (more noise). He also notes that a medium format of the image will have a similar quality of fine detail to the digital capture as these fine details are captured by the film over a larger area (yielding a higher signal to noise ratio).</p>

 

<p>The above is certainly a major factor in the perceived quality of digital images. So stepping back to the original questions I have a few more points to add.</p>

 

<p>The relationship between files size and bit depth is:</p>

 

<p><b>File Size = Number of Pixels * Bit Depth * 3 + A Number Of Bits of Overhead</b></p>

 

<p>How does the above relate with <b>photo quality</b>? There is no straight answer. The character of the capture of the light (signal to noise ratio, ...) is a major factor. But the resolution of the captured light is also a major factor. </p>

 

<p>Bit depth is just as important to direct digital capture as it is to scanning. It leaves room to "correct" a capture with far less visible numerical round off errors. Additionally, this additional color resolution will partially translate into more percieved spatial resolution if you upsize an image. Why, because smooth tones will remain smooth and will be more precise. Heck, upsizing an 8-bit capture in 16-bit mode will reduce round off errors and mildly increase the quality of an image (you might only see this when creating nose prints on a print though). </p>

 

<p>some thoughts,</p>

 

<p>Sean</p>