scanning old slides: 8 bit or 16 bit?

I'm about to embark on scanning my boxes of old Kodachrome and Ektachrome color slides. I've assumed that i should choose 16 bits/pixel to optimize the range of colors. Is there any down side to doing it in 16 bit besides the 2x larger files? Is it slower (since the time required to do this is a factor to be considered)?

I might be off, but: Bottleneck of your rig should be the cable between scanner and computer. Don't blame the wire; its the codec used on it too. File writing time should be way(!) faster than scanning time.

I have no clue how data gets transferred during the scan. - Is there an internal RAW format, using the true resolution of the scanner which should be between 8 and 16bits? Would the 8 bits get generated inside the scanner, before the reading of a pixel gets wired out? - Time to dig out your stop watch, I'd guess. Although I expect you'll see no difference.

8 bits is, plainly speaking, crap!

All the donkey work on the scan is done by the scanning software on your computer, and an 8 bit original has nowhere near enough colour or tonal data for the manipulations that need to be done - not without introducing some posterisation or loss of tonal range. (All colour or tonal image manipulations are destructive. You just can't add extra colour or tone after the fact.)

 

So don't even attempt to fart about with 8 bit scans - unless it's just for a quick 'contact sheet'.

 

A scanner isn't like a digital camera, where a lot of processing is done at 12 or 14 bit depth before spewing out an 8 bit JPEG. The scanner hardware does no manipulation on the scanned data; that's all done by your computer. So if you only feed your computer with 8 bits/channel to start with, you'll end up with maybe 6 bits/channel of useable image information.

Edited May 12, 2019 by rodeo_joe|1

As seen in the past, some of the 'advise' you're getting isn't accurate and some is utter 'crap' ;).

 

8-bits per color absolutely isn't necessarily crap and is more than enough data to output to print anything you scan. In fact, very, very few print drivers will send more than 8-bits per color to a driver anyway and the few that do (on Mac only, a couple Epson's as one example), you'll see zero difference on the print using either 8-bits of good data or 16-bits of good data.

 

High bit data (what some are incorrectly calling 16-bits in nearly every case) is about editing overhead. IF your scanner has decent quality scanning software, and of course provides high bit data to the software, you can and I'd submit in most cases should use that to get the best global tone and color and if so desired, save as either 16-bit or 8-bit per color depending on the additional IF ANY editing that scan will require. Yes, the main downside to high bit data is it's 2X the file size but storage doesn't cost much but it's a factor. Yes, your files will take longer to open and save, and process. But the idea you need 16-bit data once the image is how you desire it, and that 8-bits from that is crap is well a statement of crap. :eek:

 

Read this first, then we can move on:

http://digitaldog.net/files/TheHighBitdepthDebate.pdf

 

Next, high bit data doesn't necessarily have any more range of color than an 8-bit per color document. You're mixing up bit depth (encoding numbers) with color gamut (range of colors). An 8-bit per color document can very well have a far greater range of colors than a high bit file. In case you want to go there, and view the differences between color, color numbers and color gamut, I suggest you read this:

 

http://digitaldog.net/files/ColorNumbersColorGamut.pdf

 

Indeed, a scanner isn't like a digital camera. It produces true RGB color data but after that, bit depth between a camera and scanner follow the same rules. Someone who states a scanner does no manipulation of the data doesn't understand much about scanner software or how scanners produce color; as you desire just as if you used say Photoshop or not. Got zero to do with bit depth as any scanner worth it's salt will produce high bit data and pass it onto it's scanning software to then provide the ideal (if so set and possible depending on the tools provided) 8-bits per color. Absolutely nothing wrong or crap with that resulting data.

 

So you can scan without setting anything in high bit then try to correct that in say Photoshop. Or you can do the work at the scan stage. Or you can do a bit of both. A lot depends on the software used to drive the scanner and your workflow preferences. It doesn't take the scanner any longer to scan a 'corrected' image than one not set as such and then having to be 'fixed' later in Photoshop. Keep that fact in mind. When the rubber hits the road, you can go either way, and no, the resulting 8-bits per color scan isn't crap unless you make it crap out of ignorance.:D

 

If the scans will not be printed, then after editing, high bit isn't useful or used! Very few people have full high bit video systems in the first place and in the 2nd, most are posting JPEGs which don't support more than 8-bits per color. For a full high bit video path, the display, the video card, the OS and the software ALL have to support high bit data. If not, if one component isn't high bit, the data is converted to 8-bits per color. And for nearly all images, you'll never see a difference anyway. In fact, one has to produce a specialized document just to test IF the entire system is providing high bit video from document to display. Here's such an example:

http://digitaldog.net/files/10-bit-test-ramp.zip

I've been scanning photos for many many years--old family slides and negatives, my own film stuff, and my wife's family photos. One thing I'm sure of: get the best scan you can--highest resolution, most bits, whatever. Because 10 years from when you scan it, it will seem painfully small and crude and simple. Never fails. Technology moves so fast that today's "best" is going to be mediocre in 7=10 years. FWIW. MHO.

Standing alone, there is little if any difference between the appearance of 8 bit and 16 bit image. Where it does make is difference is when you change the color or dynamics. 8-bit files exhibit round-off errors in editing, seen as comb-tooth gaps in the histogram and possible posterization or banding in the image. Unless the editing is extreme, the histogram of a 16 bit image remains smooth. The histograms in Photoshop/Lightroom are 8-bit. 16 bit images have 8 times as many points, hence 1/8th the inevitable round-off error.

 

I agree with Ken's advice to scan at the best available resolution and bit depth. So much time is spent organizing, cleaning and scanning film, it's a shame to waste it with less than your best effort. Take time to catalog your work at the same time. You should always be able to connect each image with that roll, slide or frame number.

I'm about to embark on scanning my boxes of old Kodachrome and Ektachrome color slides. I've assumed that i should choose 16 bits/pixel to optimize the range of colors. Is there any down side to doing it in 16 bit besides the 2x larger files? Is it slower (since the time required to do this is a factor to be considered)?

 

First, the limiting factor for most, if not all scans, it the mechanical part of advancing the slide (negative) or scan head, an action measured in milliseconds. The rest of the scanning process is electrical, measured in microseconds or nanoseconds - one thousand to one million times faster.

 

If you will be editing the scans especially if you will be making color level adjustments, a higher color depth will be helpful - the more information the better.

 

Depending upon how you will be saving the scans, 8-bit may suffice. JPEG files must be 8-bit; it's part of the JPEG standard. There is no choice.

 

Me, I use a Nikon Coolscan, scan at 14-bits (the maximum for the Coolscan), edit in Photoshop, Save a PSD master file, then convert to the printer profile (in my case Costco), convert to 8-bit, and save as a JPEG. I have never been able to see any difference between the print produced by an 8-bit JPEG and a 16-bit TIFF file.

I’m not sure what hardware and software you’re using so I can only comment in hypotheticals, but...

 

If you have the ability to operate the scanner in higher than 8 bits per channel, that’s great. Pull in all the information you can, because your scanner software is going to have to do at least some work on the raw data coming from the scanner to get the exposure and color right. If you’re happy with the image produced by the software, you can save the output file in 8 bits per channel. If you’re going to do more color correction, save in higher bits.

8 bits is, plainly speaking, crap!

But (and correct):

Standing alone, there is little if any difference between the appearance of 8 bit and 16 bit image.

I have never been able to see any difference between the print produced by an 8-bit JPEG and a 16-bit TIFF file.

8 bits (per color) is, plainly speaking, crap expect when it's not and the same is true of high bit data.

The human eye can distinguish little if any more than 256 shades within a primary color or gray scale. If you print color strips with 256 steps, they will look nearly continuous to the unaided eye. All three colors combined still produce over 16000 possible shades. IMO, it is a waste of time to produce an image for print greater than 8 bits/channel, unless the customer wishes to make further adjustments or convert the image to CMYK for printing. Even so, the printing process is unlikely to contain even 8 bits/channel of information.

 

That falls apart when you make adjustments to the image, which makes the steps larger.

Hi tycin

 

Here are the decisions I have found in the scanning process

 

1) Do it yourself or pay for scanning service (you've made that decision apparently)

2) Will you be having access later on to the slides for the rare case you want a better scan? If yes, you can scan at lower resolution and to JPEG and for the rare few, go back and do a better scan later. This saves on time and disk space yet personally I don't want to have to save my slides in safe storage forever yet it is a viable option.

2) Which resolution to scan at (which does impact scanning speed and file size) determined by

- what's the largest print you want to ever make

- what is the smallest area on a slide that you might want cropped out to print

3) How much post processing to you expect on these scans. If significant adjustments needed in recovery work, scanning at higher resolution, 16 bit, and Tiff format has proven useful for me. I suspect in the majority of old family images, that would not be the case (see #2 above). If not much post processing then 8 bit scans to good quality JPEG should be just fine. So again, option #2 could work for just the slides that need that special processing

 

Just some quick thoughts to consider based on my own experience. Best wishes on your project.

Edited May 12, 2019 by john_wheeler|6

Dog, we're talking about the raw information from the scanner being 8 bit, not the output to the printer or after colour correction. Edited May 14, 2019 by Sandy Vongries

As seen in the past, some of the 'advise' you're getting isn't accurate and some is utter 'crap' ;).

 

8-bits per color absolutely isn't necessarily crap and is more than enough data to output to print anything you scan. In fact, very, very few print drivers will send more than 8-bits per color to a driver anyway and the few that do (on Mac only, a couple Epson's as one example), you'll see zero difference on the print using either 8-bits of good data or 16-bits of good data.

 

High bit data (what some are incorrectly calling 16-bits in nearly every case) is about editing overhead. IF your scanner has decent quality scanning software, and of course provides high bit data to the software, you can and I'd submit in most cases should use that to get the best global tone and color and if so desired, save as either 16-bit or 8-bit per color depending on the additional IF ANY editing that scan will require. Yes, the main downside to high bit data is it's 2X the file size but storage doesn't cost much but it's a factor. Yes, your files will take longer to open and save, and process. But the idea you need 16-bit data once the image is how you desire it, and that 8-bits from that is crap is well a statement of crap. :eek:

 

Read this first, then we can move on:

http://digitaldog.net/files/TheHighBitdepthDebate.pdf

 

Next, high bit data doesn't necessarily have any more range of color than an 8-bit per color document. You're mixing up bit depth (encoding numbers) with color gamut (range of colors). An 8-bit per color document can very well have a far greater range of colors than a high bit file. In case you want to go there, and view the differences between color, color numbers and color gamut, I suggest you read this:

 

http://digitaldog.net/files/ColorNumbersColorGamut.pdf

 

Indeed, a scanner isn't like a digital camera. It produces true RGB color data but after that, bit depth between a camera and scanner follow the same rules. Someone who states a scanner does no manipulation of the data doesn't understand much about scanner software or how scanners produce color; as you desire just as if you used say Photoshop or not. Got zero to do with bit depth as any scanner worth it's salt will produce high bit data and pass it onto it's scanning software to then provide the ideal (if so set and possible depending on the tools provided) 8-bits per color. Absolutely nothing wrong or crap with that resulting data.

 

So you can scan without setting anything in high bit then try to correct that in say Photoshop. Or you can do the work at the scan stage. Or you can do a bit of both. A lot depends on the software used to drive the scanner and your workflow preferences. It doesn't take the scanner any longer to scan a 'corrected' image than one not set as such and then having to be 'fixed' later in Photoshop. Keep that fact in mind. When the rubber hits the road, you can go either way, and no, the resulting 8-bits per color scan isn't crap unless you make it crap out of ignorance.:D

 

If the scans will not be printed, then after editing, high bit isn't useful or used! Very few people have full high bit video systems in the first place and in the 2nd, most are posting JPEGs which don't support more than 8-bits per color. For a full high bit video path, the display, the video card, the OS and the software ALL have to support high bit data. If not, if one component isn't high bit, the data is converted to 8-bits per color. And for nearly all images, you'll never see a difference anyway. In fact, one has to produce a specialized document just to test IF the entire system is providing high bit video from document to display. Here's such an example:

http://digitaldog.net/files/10-bit-test-ramp.zip

 

8 Bit is enough for printing or displaying but if you need to edit then it's not enough. So I would scan as 16 bit. Making edit and finally save as 8 bit JPEG is fine.

Dog, we're talking about the raw information from the scanner being 8 bit, not the output to the printer or after colour correction.

 

 

It's not raw nor 8 bits per color.

Edited May 14, 2019 by Sandy Vongries

The incremental time it takes to scan at high resolution and bit depth is negligible compared to the time it takes to locate, clean and rescan a slide or film strip later. Time may not be important now, but it will be later ;)

The incremental time it takes to scan at high resolution and bit depth is negligible compared to the time it takes to locate, clean and rescan a slide or film strip later. Time may not be important now, but it will be later ;)

Indeed and as I mentioned, do as much tonal and color 'correction' as possible at the scan stage instead of on full resolution data after in say Photoshop. That's if (a big if), the scanning software is good. Even if you're scanning in high bit, it can be a better use of time to apply those corrections at the scan stage; less work to do later. It's why really good scanning software is so important, otherwise the scanning hardware is an expensive door stop.

 

SilverFast is pretty powerful. The GUI isn't tops for being intuitive IMHO but there's little it can't provide. My favorite scanning software from the past was LinoColor which had some functionality that put (at the time) Photoshop to shame. PS still doesn’t have a saturation curve like LinoColor did. ColorQuartet which drove one of my drum scanners was also very good, allow nearly all corrections at the scan stage. And of course, handing off high bit data after the scan for additional editing if necessary or simply to provide all the data for future needs. Today, there's zero reason to send high bit to a print driver, even Epson admitted this when I asked about the option in their 3880 driver. But they said in the future, high bit could be useful for print output. But again, high bit today is about editing overhead such we send the best 8-bits per color to whatever output we desire.

Despite all the "enlightenment" above.

 

It is simple

"Scan high, scan once"

 

The limits on how much to preserve do run up against the amount of storage you have, but a resolution much above 4000 ppi will rarely yield more actual data. Bit-Depth is another place where the original slide sets limits on what you can get from any given slide or image.