How many bits per color are needed

One of the questions that gets debated from time to time is how many
bits / color are needed for an image. The debate is usually between 8
and 16 bit per color with the argument that 8 bits is not enough to
handle the shadow part of the image. I got playing around with this
and under the right conditions it is pretty amazing just how good the
shadow detail can be using 8 bits.

Whereas 8bits/color is enough for great shadow detail the jpeg file
format tends to trash the dark parts of the image. If you want to have
very good shadow detail a lossless format such as tiff is needed.

This is an image where the whole of it is jammed into the bottom 32
levels.
http://www.sewcon.com/temp/shadow.tif
This of course looks very dark but using levels the detail can be
extracted to look like this.
http://www.sewcon.com/temp/shadow%20expanded.tif

Trying to store this same image as a jpeg and expanding it does not
work nearly as well.
http://www.sewcon.com/temp/jpg_shadow_expanded.jpg

I should point out that the only reason you would need this level of
detail in the shadows is if you are going to be making some pretty
extreme adjustments to the photo.

Scott

That is truly amazing. I hadn't realized how much detail was there.

I just downloaded your shadow.tif and used the GIMP's "curves"
control to bring out the detail and, sure enough, it was all right
there.

In case anyone thinks Scott was cheating, he wasn't. The beautiful
image in "shadow expanded.tif" really is hiding in shadow.tif.

Alan

Scott W wrote:

One of the questions that gets debated from time to time is how many
bits / color are needed for an image. The debate is usually between 8
and 16 bit per color with the argument that 8 bits is not enough to
handle the shadow part of the image. I got playing around with this
and under the right conditions it is pretty amazing just how good the
shadow detail can be using 8 bits.

Whereas 8bits/color is enough for great shadow detail the jpeg file
format tends to trash the dark parts of the image. If you want to have
very good shadow detail a lossless format such as tiff is needed.

This is an image where the whole of it is jammed into the bottom 32
levels.
http://www.sewcon.com/temp/shadow.tif
This of course looks very dark but using levels the detail can be
extracted to look like this.
http://www.sewcon.com/temp/shadow%20expanded.tif

Trying to store this same image as a jpeg and expanding it does not
work nearly as well.
http://www.sewcon.com/temp/jpg_shadow_expanded.jpg

I should point out that the only reason you would need this level of
detail in the shadows is if you are going to be making some pretty
extreme adjustments to the photo.

Scott


Expand first and then save as a jpeg.

Scott W wrote:
One of the questions that gets debated from time to time is how many
bits / color are needed for an image. The debate is usually between 8
and 16 bit per color with the argument that 8 bits is not enough to
handle the shadow part of the image. I got playing around with this
and under the right conditions it is pretty amazing just how good the
shadow detail can be using 8 bits.

Whereas 8bits/color is enough for great shadow detail the jpeg file
format tends to trash the dark parts of the image. If you want to have
very good shadow detail a lossless format such as tiff is needed.

This is an image where the whole of it is jammed into the bottom 32
levels.
http://www.sewcon.com/temp/shadow.tif
This of course looks very dark but using levels the detail can be
extracted to look like this.
http://www.sewcon.com/temp/shadow%20expanded.tif

Trying to store this same image as a jpeg and expanding it does not
work nearly as well.
http://www.sewcon.com/temp/jpg_shadow_expanded.jpg

I should point out that the only reason you would need this level of
detail in the shadows is if you are going to be making some pretty
extreme adjustments to the photo.

Scott



Room for adjustments is one reason. Changing conversion to CMYK for
another type of printing method can be another. Nearly all commercial
printing and layout software only handles 8 bit images. It could be
argued that an 8 bit workflow could save some time. At least with
scanning, most scanners are faster handling an 8 bit scan. Another
aspect of this to consider is that some scanners (and some digital
cameras) are actually only 10 bit or 12 bit, even though in PhotoShop
anything more than 8 bits is worked in 16 bit mode. Given a 10 bit
scanner or digital camera, there might not be much benefit for going to
some printing types.

Ciao!

Gordon Moat
A G Studio
http://www.allgstudio.com

Scott W wrote:
One of the questions that gets debated from time to time is how many
bits / color are needed for an image. The debate is usually between 8
and 16 bit per color with the argument that 8 bits is not enough to
handle the shadow part of the image.

Heh. It also depends wether the colour space
is linear or logarithmic.

BugBear

bugbear wrote:
Scott W wrote:

One of the questions that gets debated from time to time is how many
bits / color are needed for an image. The debate is usually between 8
and 16 bit per color with the argument that 8 bits is not enough to
handle the shadow part of the image.


Heh. It also depends wether the colour space
is linear or logarithmic.

BugBear

Why???????

Why be a smart ass unless you're prepared to share your knowledge?

Or have I answered that question myself?

What you are really talking about here is exposure, not color.

Scott W wrote:
One of the questions that gets debated from time to time is how many
bits / color are needed for an image. The debate is usually between 8
and 16 bit per color with the argument that 8 bits is not enough to
handle the shadow part of the image. I got playing around with this
and under the right conditions it is pretty amazing just how good the
shadow detail can be using 8 bits.

Whereas 8bits/color is enough for great shadow detail the jpeg file
format tends to trash the dark parts of the image. If you want to have
very good shadow detail a lossless format such as tiff is needed.

This is an image where the whole of it is jammed into the bottom 32
levels.
http://www.sewcon.com/temp/shadow.tif
This of course looks very dark but using levels the detail can be
extracted to look like this.
http://www.sewcon.com/temp/shadow%20expanded.tif

Trying to store this same image as a jpeg and expanding it does not
work nearly as well.
http://www.sewcon.com/temp/jpg_shadow_expanded.jpg

That's because the JPEG quantisation figures
used in "Normal" desktop applications
are tuned for natural (real world) images
with "normal" colour distributions.
JPEG (in the abstract) could handle your example
just fine, but you'd need a JPEG guru to work
out good quantization figures.

(and I don't just mean thr "quality" number)
http://www.impulseadventure.com/phot...ntization.html

BugBear

"John A. Stovall" wrote in message
...
On Wed, 01 Feb 2006 02:11:59 GMT, no_name
wrote:

Scott W wrote:

One of the questions that gets debated from time to time is how many
bits / color are needed for an image. The debate is usually between 8
and 16 bit per color with the argument that 8 bits is not enough to
handle the shadow part of the image. I got playing around with this
and under the right conditions it is pretty amazing just how good the
shadow detail can be using 8 bits.

Whereas 8bits/color is enough for great shadow detail the jpeg file
format tends to trash the dark parts of the image. If you want to have
very good shadow detail a lossless format such as tiff is needed.

This is an image where the whole of it is jammed into the bottom 32
levels.
http://www.sewcon.com/temp/shadow.tif
This of course looks very dark but using levels the detail can be
extracted to look like this.
http://www.sewcon.com/temp/shadow%20expanded.tif

Trying to store this same image as a jpeg and expanding it does not
work nearly as well.
http://www.sewcon.com/temp/jpg_shadow_expanded.jpg

I can't save either one of these images. I get a message that says, "buy
Quick time", and when I click on OK, I'll buy" nothing happens......

This thread is recognizing the differences between implimentations of jpeg compression on different cameras. My kodak DX 7590 has two different jpeg compression levels and no raw the best (most detailed image) jpeg compression removes a lot of detail
from the image including brightness information on bright objects that appears in the electronic viewfinder not not in the stored image. An old small canon S110 that I have uses a much better jpeg compression implementation and seems to preserve the
dynamic range. Both camera's have 8 bit color levels.

To get back to the real thread for general use 8 bits are enough in the same way that the S110's 2M pixels is enough. Properly framed so that cropping is not needed 2M pixels produces nice images.

Properly exposed so post processing light levels are not needed 8 bits is enough for some very good images. If recovering washed out detail or if brightness/contrast is an issue then 8 bits are not enough. The next digital camera I will want more color
levels over more pixels.

w..


Scott W wrote:

One of the questions that gets debated from time to time is how many
bits / color are needed for an image. The debate is usually between 8
and 16 bit per color with the argument that 8 bits is not enough to
handle the shadow part of the image. I got playing around with this
and under the right conditions it is pretty amazing just how good the
shadow detail can be using 8 bits.

Whereas 8bits/color is enough for great shadow detail the jpeg file
format tends to trash the dark parts of the image.