TIFF to DNG via lightroom

[stuart_stent]

I did some test scans last night using vuescan to TIFFs(48bit).

 

The result was a lovely 60mb tiff.

 

I then imported these tiffs into lightroom and converted them to DNG. Resulting

in a 16mb DNG.

 

my question is, am I losing any data by doing this, or is the DNG just a lot

more efficient?

[seland]

My guess: The DNG applies a lossless compression, like LZH for TIFF.

[mendel_leisk]

*If* there is no loss that is one excellent compression tool. Hmmm: can you convert the DNG's back to tiff with LightRoom and produce a tiff with identical histogram statistics in Photoshop? I forget all the names, one's Median, and there are at least a couple more. Can you instruct LightRoom to avoid any auto-adjustments in the conversion back to tiff?

[marc_rochkind]

Sounds normal to me.

 

I produced some TIFFs from Vuescan, and they were uncompressed. It's certainly possible that compression could reduce the size by that much. I've seen some of my files go from 35MB to 11MB.

 

There's no harm, as the compression in the DNG file is lossless. I see that Vuescan has an option for compression... apparently I didn't use it.

[mendel_leisk]

Marc, re: "I've seen some of my files go from 35MB to 11MB", doing what, what file format, what steps?

 

And re: "It's certainly possible that compression could reduce the size by that much.", I think going from 60 to 16, *if* it's lossless, is bloody amazing. It is better by far than lossless compressed jpeg2000. With (grainy) scan files I got maybe 40% reduction with jpeg2000, starting with 16bit tiffs, but then found Photsohop was unable to open them, and ultimately decided this was not something I wanted to risk.

 

Per my first response, Tiff>DNG>Tiff is only worthwhile for compression if you can restore the tiff from the DNG and have a pixel-by-pixel copy of the original. The histogram statistics should be identical.

 

Also, regarding Vuescan's option for compression, it's the one program I've found that was able to implement tiff compression that actually reduced the file size, albeit not by much. I was getting 10~20% compression with grainy scans. I decided it wasn't significant enough to use: The size reduction wasn't that much, it made for slower loading files, and I didn't want to risk it.

 

Photoshop, by comparison, made tiff files *bigger* with lzw compression.

[marc_rochkind]

Mendel--

 

I don't have all the info at hand, but the DNGs in question were produced by opening the TIFFs in Lightroom and then immediately saving them as DNGs.

 

The ratio is not only dependent on the denominator (the DNG). It's also dependent on the numerator: The amount of air in the original Vuescan TIFFs.

 

Restoring the TIFF was not important to me, as all my processing nowadays goes through Lightroom, and then, for some images, into Photoshop.

[mendel_leisk]

I'm also interested in what I could do with my tiffs using a raw processor, but my first interest would be file compression. I have large film scan tiffs, each around 226,000KB. For an exercise, I tried a DNG conversion through Adobe Bridge and Adobe Camera Raw. I chose lossless compression, and specified for the conversion to be linear. I got nowhere near the OP's compression ratio (largely due to this being a grainy scan file I would think), but still not bad:

 

226730 tiff > 166475 dng

 

The image I tried had corner to corner detail and was from a moderately overexposed slide. Maybe with a more usual image, the compression would improve. Still, not bad, I think about the same as jpeg2000.

 

Still with ACR, I resaved the DNG back to a new tiff. The resulting tiff was slightly different in size: 226739 (maybe due to the introduced "fly" ;) )

 

Opening both old and new tiffs in photoshop, they appear identical, the histograms appear identical, and the histogram statistics mean, std dev and median are all identical. This is after "uncaching" both files histograms.