Known Participant

질문

Lightroom handling of 16 Bit vs 8 Bit images

Forum|Forum|7 years ago
May 2, 2018
6 답변들
12386 조회

I have a few thousand scans of 35 mm transparancies.

They need some gentle editing / cropping but I want to preserve the masters as well as any edits.

Only trouble is, each file is 233Mb - the scans are 16 bit, 5400 dpi TIFF files.

So, while I want to preserve the masters without any edits, I really could do with decreasing the file sizes.

I can save to ZIPTIFF but actually that only saves about 10%.

So to the real question – is 16-bit (vs 8 bit) actually giving me anything? Making them 8-bit TIFF files would really save on filesize.

I ask because I have converted a few in Lightroom to 8 bit just to see and I can’t tell ANY difference between 8 and 16 bit on screen – not what I expected.

Is it because I am not looking at them correctly (or in the correct application)?

Any pointers would be great – thanks!

이 주제는 답변이 닫혔습니다.

99jon

Legend

I export my scanned tiffs as Adobe DNG. This usually produces a significant reduction in file size without loss of quality.

The DNG's become the lr working copies after adding to the catalog. I then remove and archive the tiff files to an external drive.

Jao vdL

Community Expert

I would actually add a different caution. Very few scanners have a low-enough noise floor that you need a full 16 bits to accurately store the data from the scanner. That is apart from the fact that the film itself generally does not have a good enough signal to noise ratio to warrant 16 bits even if the scanner says it is designed to give you 16 bits per channel. The 16-bits number is the resolution of the ADC converter used but the noise in your scan is actually determined by the lamp, the film density/thickness, whether you scan a negative or a positive and the nature (grain) of the film and how you exposed it and developed it. There are some subtleties with whether you scanned negatives (generally there it is even worse) or positives and what gamma curve you used in the file that can change this a bit in limited parts of the image but generally 16-bit files from most scanned film images are overkill. The fact that the zip compression only gives you a 10% reduction of file size is actually a good indication that many of the bits in the file are noise. You generally won't see that by just looking at the image on your screen but it is quite indicative. Noise is almost impossible to compress and so if your file is noisy you'll end up not being able to compress the image much.

That said, I generally do the same thing and still store my master scans as high res high bit depth simply because disk space is cheap.

Todd Shaner

Legend

https://forums.adobe.com/people/Jao+vdL wrote
I would actually add a different caution. Very few scanners have a low-enough noise floor that you need a full 16 bits to accurately store the data from the scanner. That is apart from the fact that the film itself generally does not have a good enough signal to noise ratio to warrant 16 bits even if the scanner says it is designed to give you 16 bits per channel.

But what happens to the 8 bit image data after applying toning edits. Here are histograms of a Kodachrome 64 slide captured with a Plustek 7600i film scanner. Its Dmax is claimed to be a modest 3.5 and Kodachrome has one of the highest Dmax of any film type so it fits your example. A simple PS Curves adjustment was applied to correct Kodachrome's high contrast and dense shadow areas and the Black and White clipping points set. As bad as this looks it's not visible in the image at 1:1 view, but what about images that need more extreme toning and color corrections? Interestingly the 8 bit TIFF doesn't exhibit the comb pattern in the LR's Histogram. This is due to LR creating a less accurate histogram from a down-sampled copy of the Develop image bit map. You can see the same effect in PS's histogram when it's using the cached image. A small triangle with ! is displayed in the Histogram panel and the 8 bit and 16 bit histograms look identical same as in LR, but it's not the actual image data histogram. In PS you can click on the 'uncached refresh' button to update the Histogram and make it accurate. Don't be fooled by LR's Histogram!

16 bit TIFF With Curve Adjustment 8 bit TIFF With Curve Adjustment

Todd Shaner

Legend

The combed histogram doesn't actually necessarily indicate that the 8-bit tiff after curve adjustments has lower image quality. You might just be stretching noise and you won't be able to see it. This is easy to illustrate by generating a noisy gradient in Photoshop and adding a bit of noise (1% in this case). Then convert a copy of this to 8 bits and apply the same curves adjustment to both images. Here is the curves adjustment:

Here are the resulting histograms (16 bits left and 8 bits right):

You can see that the 8 bits image is all chopped up in the histogram. But look at the images:

They are identical and both equally non-posterized even though one histogram looks terrible. The reason is that the noise is acting like dithering which will hide any possible posterization. If I do this to a non-noisy gradient you do see posterization in the 8-bits image as you can see below for same gradient, same curves but no noise in the original. The histogram on this is basically identical to the 8 bits histogram above but you can clearly see posterization all through!

So if you have noise in your image that surmounts the 8th bit of the image you can safely use 8bits even if your histogram will look ugly after doing strong curves manipulation. The image will be visually identical. I used a 40dB signal to noise ratio above which is very typical for flatbed scanners. You can do much better with a drum scan which uses a photomultiplier tube instead of a ccd to record the image and has a far better signal to noise ratio. For drum scans, go 16 bits!

Jao, you're probably correct concerning most scanned film images. The 35mm Kodachrome slide I used in the example exhibits no banding in the 8 bit TIFF file with edits applied. I just wanted to make WPA57 aware of the possibility for that happening. We don't know what type of scanner was used or if grain/noise reduction was applied using the scanner's software. If I was still using a film scanner that took minutes per film image there's no way I'd choose 8 bit. I now use DSLR film copier setup with a 21 megapixel raw image file about 28 MB (14 bit) versus 16 bit TIFF 126 MB. It also way faster! That won't help the OP with the task at hand, but a thought for future projects.

WPA57작성자

Known Participant

Many thanks to you all for your input here.

Taking everything together, I think I can summarise:-

- If the amount of editing I need to do to get to my final result is low, then I can probably get away with the masters being downgraded to 8-bit.

- to the question Were you comparing them before or after applying edits? If you imported them as 16 bpc and exported them as 8 bpc without editing, any differences might not be visible. The differences might still be hard to see if any edits were minor. Again, where an 8 bpc starts to show visible deterioration is after major color or tone edits are applied.

Yes I was comparing without ANY edit, just 8 bit vs 16 bit versions of the same file.

- I can also reduce the dpi (It's also an interesting comment about scanning high then downgrading being better than scanning at the lower resolution)

2 Questions remaining

1/ Why would I need to use Photoshop (Elements?) not Lightroom to downgrade to 8-bit? (also - will Lightroom handle dpi reduction?)

2/ Is ZIP or LZW still the most effective lossless compression? (btw I confirm that LZW on 16 bit increases filsize!)

Per Berntsen

Community Expert

1/ Why would I need to use Photoshop (Elements?) not Lightroom to downgrade to 8-bit? (also - will Lightroom handle dpi reduction?)

Lightroom cannot convert a file from 16-bit to 8-bit, because it never changes the original. (the image you see on screen is a preview created by Lightroom) All the edits you do in Lightroom are parametric - i.e. they are written as text to the catalog, and they are applied to the preview, but not to the original.

When you export a file, a copy of the file is created, and the edits you have done are applied to the new file, and in the export dialog you can resize the file, and also set the bit depth to 8-bit or 16.bit.

So if you want to convert all your tiffs to 8-bit, the best way to do this is to use Photoshop.

You can create an action to do this, including using LZW compression, and run the action on the folders with the tiffs.

Like I said before, I would back up all the (16-bit) files first, and at least keep them until you are sure that there has been no problems with the conversion. Once you have converted, you can't convert back to 16-bit. (well you can, but it's pointless,since all the 16-bit info will be gone, and there will no improvement in quality)

2/ Is ZIP or LZW still the most effective lossless compression? (btw I confirm that LZW on 16 bit increases filsize!)

ZIP is most effective with 16-bit, LZW is most effective with 8-bit.

A few words about dpi and ppi.

Dpi means dots per inch, and is used exclusively for printers. It describes how many ink dots the printer can print per inch.

Ppi means pixels per inch, and is also used for printing, but it describes how many image pixels are used to print one inch.

There is a relationship between pixel dimensions, printed dimension and ppi - if you want to print an 8 x 10" image at 300 ppi, you will need a certain number of pixels in the image. How many? 2400 x 3000, because 8x300 = 2400 and 10x300 = 3000.

Ppi is also used to describe the resolution of a scanner (some people prefer to use spi - samples per inch - each sample becomes an image pixel).

You know that your scanner has a resolution of 5400 ppi, and you can then calculate the pixel dimensions.

You are scanning 35mm film, which is roughly 1 x 1½", and when you multiply the ppi with the dimensions, you get 5400 x 8100 pixels. If you were to print the full image at 300 ppi, the print size would be 18 (5400:300) by 27 (8100:300) inches.

WPA57작성자

Known Participant

https://forums.adobe.com/people/Per+Berntsen wrote
1/ Why would I need to use Photoshop (Elements?) not Lightroom to downgrade to 8-bit? (also - will Lightroom handle dpi reduction?)
Lightroom cannot convert a file from 16-bit to 8-bit, because it never changes the original. (the image you see on screen is a preview created by Lightroom) All the edits you do in Lightroom are parametric - i.e. they are written as text to the catalog, and they are applied to the preview, but not to the original.
When you export a file, a copy of the file is created, and the edits you have done are applied to the new file, and in the export dialog you can resize the file, and also set the bit depth to 8-bit or 16.bit.
So if you want to convert all your tiffs to 8-bit, the best way to do this is to use Photoshop.
You can create an action to do this, including using LZW compression, and run the action on the folders with the tiffs.
Like I said before, I would back up all the (16-bit) files first, and at least keep them until you are sure that there has been no problems with the conversion. Once you have converted, you can't convert back to 16-bit. (well you can, but it's pointless,since all the 16-bit info will be gone, and there will no improvement in quality)
2/ Is ZIP or LZW still the most effective lossless compression? (btw I confirm that LZW on 16 bit increases filsize!)
ZIP is most effective with 16-bit, LZW is most effective with 8-bit.
A few words about dpi and ppi.
Dpi means dots per inch, and is used exclusively for printers. It describes how many ink dots the printer can print per inch.
Ppi means pixels per inch, and is also used for printing, but it describes how many image pixels are used to print one inch.
There is a relationship between pixel dimensions, printed dimension and ppi - if you want to print an 8 x 10" image at 300 ppi, you will need a certain number of pixels in the image. How many? 2400 x 3000, because 8x300 = 2400 and 10x300 = 3000.
Ppi is also used to describe the resolution of a scanner (some people prefer to use spi - samples per inch - each sample becomes an image pixel).
You know that your scanner has a resolution of 5400 ppi, and you can then calculate the pixel dimensions.
You are scanning 35mm film, which is roughly 1 x 1½", and when you multiply the ppi with the dimensions, you get 5400 x 8100 pixels. If you were to print the full image at 300 ppi, the print size would be 18 (5400:300) by 27 (8100:300) inches.

Good points, thankyou - and yes I get what you mean about dpi and ppi.

Per Berntsen

Community Expert

Scanning at the max resolution of the scanner is recommended procedure, since downsizing a full resolution file will give you better quality than scanning at a reduced ppi. You could of course downsize the files in Photoshop before importing them, but you would be throwing away information. And although the file size is quite large, they are only about 39 megapixels.

dj_paige is right about the limitations of 8-bit editing, but this only applies to editing pixels in Photoshop.

Editing in Lightroom is different, because the edits are only applied when you export, and then to an entirely new file - the original remains untouched. IOW, there is not the same potential for a lower quality file as when you edit in Photoshop.

So you could consider converting the files to 8-bit in Photoshop, and you could also use LZW compression, which is more effective than ZIP compression is with 16-bit files. (LZW compression is inefficient with 16-bit files, and can in fact increase the file size) Personally, I would first save copies of the 16-bit files to an external drive - I would never delete a 16-bit version of a file. If you need do to heavy editing in Photoshop (things you can't do in Lightroom), 16-bit is required.

Conrad_C

Community Expert

If the original scans are pretty close to how they should look, so that any tone and color adjustments are going to be minor, then you might get away with storing them as 8 bpc TIFF.

But if the original scans need any big color or tone moves, especially something like pushing the Shadows slider up a lot, it's probably risky to have 8bpc originals.

WPA57 wrote
I ask because I have converted a few in Lightroom to 8 bit just to see and I can’t tell ANY difference between 8 and 16 bit on screen – not what I expected.
Is it because I am not looking at them correctly (or in the correct application)?

Were you comparing them before or after applying edits? If you imported them as 16 bpc and exported them as 8 bpc without editing, any differences might not be visible. The differences might still be hard to see if any edits were minor. Again, where an 8 bpc starts to show visible deterioration is after major color or tone edits are applied.

As far as software, Lightroom is a good place to edit 8 bpc images because it only applies edits to the exported or printed version, so that edits are applied as few times as possible (maybe only once). An 8 bpc image will visually deteriorate much faster if multiple rounds of edits are applied to it, which is what would happen if it was edited in Photoshop without adjustment layers.

dj_paige

Legend

screens are only 8 bits, you can't see the difference between 16 bits and 8 bits on your screen.

However when doing the calculations needed to edit, a 16 bit original will produce higher quality edits than the same edits on an 8 bit original, and this you can see on an 8bit monitor when you are finished editing.

So I think you are throwing away useful and important information by going to 16 bits to 8 bit originals.

The place where you might be able to save space is that you almost definitely don't need to scan photos at 5400 dpi. Perhaps that's the place a reduction can be made, without losing noticeably quality. Unfortunately, if the images have been scanned, I'm not sure there's an easy way to lower this, you could possibly import the photos into Lightroom and then reduce the resolution.

가입하기

소셜 로그인

커뮤니티에 로그인

소셜 로그인

파일을 바이러스 검사 중입니다.

이 파일은 다운로드할 수 없습니다