Peabody <waybackNO784SPA...@yahoo.com> wrote: >I understand that raw files are typically 12-bit or 14-bit, instead >of the 8-bit found in jpegs. But when you are editing raw images, >you don't actually see the 12-bit image do you? I mean, insn't the >display going to show you an 8-bit picture? So every time you make >a change to something in the image, the software has to re-render >the 12-bit info into an 8-bit version for the display.
>Or do I have this all wrong?
Well... the RAW data is *never* what you look at anyway! It necessarily has to be interpolated to get an image format. The resulting image might be a JPEG, but it could be a PPM or a TIF format too, and it could just as easily also be the internal format for any one of a number of image editing programs too. Regardless of what it is though, you are correct that it must be converted to an RGB format to view it... and internally some monitors are necessarily going to convert it to a 6-bit format!
On Sat, 07 Nov 2009 22:50:12 -0600, Peabody <waybackNO784SPA...@yahoo.com> wrote:
>I understand that raw files are typically 12-bit or 14-bit, instead >of the 8-bit found in jpegs. But when you are editing raw images, >you don't actually see the 12-bit image do you? I mean, insn't the >display going to show you an 8-bit picture? So every time you make >a change to something in the image, the software has to re-render >the 12-bit info into an 8-bit version for the display.
>Or do I have this all wrong?
Correct. (or as someone else already noted, to bit-levels even less than 8) When printed, reducing this even further.
12-bit and 14-bit RAW is popularized by "photographers" who must greatly expand some tonal range in their image during editing, without introducing large gaps in the recorded colors and intensities. Mostly because they are nothing but lowly snapshooters dependent on auto-exposure settings, people that don't know how to correctly expose an image in the first place. Or because their camera is really poor at doing its own RAW to JPG conversion--in the process it will clip most of the useful ends of the dynamic range of the sensor (better, properly programmed cameras do not do this). Again, proving that they are unskilled photographers because they don't even know how to purchase decent cameras in the first place.
This extended bit-range is important for them in editing because:
(Using exaggerated decimal values of an 8-bit tonal range, for clarity.) You might have an important area of your image that is highly underexposed (from poor photography skills). A large and important area of your scene being recorded in values between 0 and 19 (out of 255 possible 8-bit levels). In order to make that image the least bit useful the "photographer" will try to expand that 0 to 19 value range of details to 0 to 99 in their editor to try to get the details more visible out of those greatly underexposed areas. You would now only have 20 values representing almost half (2/5ths) of the image's available hues and intensities. Huge tonal gaps represented by posterized effects (bands of solid hues) in the image. Ghastly looking unless used for special artistic effects (known as posterizing).
This is why they applaud 12-bit and 14-bit RAW files so much. To help compensate later for their total lack of skills at photography (wrong exposures, wrong white-balance settings, ... on ad-infinauseum). Or in not knowing how to purchase decent cameras that don't require access to the RAW data, because the full sensor range is already represented in its properly exposed JPG files.
Those who preach and worship their God-RAW only reveal their total lack of skill and talent.
> > Well... the RAW data is *never* what you look at > > anyway! It necessarily has to be interpolated to get an > > image format.
>Let me ask about that too. As I understand it, you've got >this Bayer pattern screen built into the sensor, so that >each pixel is really only registering one color. So at some >point, there has to be the interpolation to create for that >pixel what the other two colors probably would have been - >sort of a synthetic Foveon situation.
>Has that already been done by the camera in the process of >producing the raw file, or is the raw file really raw and >therefore contains only one color per pixel?
They are usually referred to as photosites on the sensor (some calling them sensels, as a play on sensor+pixels), later interpolated into RBG image pixels.
Each photosite carries a numeric value, no real color. Its number is assigned a color value by its position on the sensor, being under one of 4 R, G, G, B filters in the Bayer array.
> If the latter, >then I assume there's something in the raw file format that >tells the editing software what the color pattern is. And >are there different ways to do the interpolation, so that >you get different results from different software before you >even begin the editing?
During importing the RAW sensor data into the editor it must undergo the transformation to an RBG image in order for you to view it. There are many many RGGB, GBRG, etc. orders of patterns on various sensors. Some even having 4 different colors instead of just 3 colors in a group of 4 (2 photosites per each red and blue, typical Bayer sensor).
If you would like to test various import interpolation methods to see what they do with the RGGB data converted to RGB data, download Photoline from www.pl32.net and play with it during the 30 day demo period. It has 6 different RAW interpolation methods to choose from, to show you how some are more efficient or accurate than others. (Under: File > Import > Digital Camera RAW Data) Point it at a RAW file to open, then on the left side of the Raw Import preview pane is a drop-down list of various interpolation methods. I most often use the "VNG Interpolation Cubic" option, but depending on the type of fine details in an image, others might be better at times. Zoom in on the image so you can see the individual RGB pixels to see how the interpolation changes with each type selected. If you have a Gretag-Macbeth color chart to photograph, even one found on some web-page and displayed on your monitor screen, that's a good test to see how the interpolations handle various colors and dithering patterns for them. Yellows and oranges seem to be problematic for some RAW interpolation methods. If photographing a Gretag-Macbeth chart displayed on your monitor, photograph it with the lens defocused so the screen's visible pixels don't compound the issue, unless you'd like to see what the interpolation methods also do to fine moire' causing details.
(Before opening any RAW files in Photoline, be sure to go to Edit > Options
> File > Raw, and set the gamma to your workspace, 2.20 if on a PC,
otherwise any RAW images will appear far too dark. Also set the Automatic White-Point to "on" if you are using CHDK RAW files, as they don't contain the white-point information in the RAW header.)
> > Regardless of what it is though, you are correct that it > > must be converted to an RGB format to view it... and > > internally some monitors are necessarily going to > > convert it to a 6-bit format!
>Yeah, I suspect my cheap Toshiba laptop is one of those. >Well, I guess it's just a reminder that you just don't know >what your pictures are going to look like to those seeing >them on a computer.
>Anyway, I just wanted to be clear that you dont actually see >those 14 bits of granularity while you're editing the >picture. You get the benefit of them being there when >making adjustments, but at best you are actually looking >at an 8-bit rendition of the picture.
>In my case it's only 10-bit raw, and I'm just not sure >that's worth the bother.
It depends on your purpose and needs. If you are referring to a CHDK RAW data file (many of them being 10-bit), then you can tweak out slightly more resolution from the RAW file by using more advanced RAW interpolation methods and noise-reduction tools than what the camera uses to create its own JPG file. (Every other year or so someone comes up with a slightly better RAW interpolation mouse-trap.) The increase in resolution by using more advanced methods than the camera uses on the RAW data is slight, but it is observable by photographing resolution test charts or detailed foliage. Those camera are also very good at accurately depicting the full dynamic range of the sensor in the JPG file. But some have found that color information at the extremes of the sensor's dynamic range is more accurately retained in the RAW data. Where a dark red carpet in deep shadows might look nearly black in the JPG file, by using the RAW data it will now appear to have a more reddish hue. In either case, the improvement is slight because of how well those cameras already interpret the RAW data for their JPG images. For some subjects you might find that little bit extra important, so use RAW. You'll have to compare and test it yourself to see if RAW is even worth the bother for your style and requirements of photography. I personally find only about 3 times per month where RAW from my CHDK capable camera is worth the extra steps required for special subjects. The JPGs from them are already really good.
> > Well... the RAW data is *never* what you look at > > anyway! It necessarily has to be interpolated to get an > > image format.
>Let me ask about that too. As I understand it, you've got >this Bayer pattern screen built into the sensor, so that >each pixel is really only registering one color. So at some >point, there has to be the interpolation to create for that >pixel what the other two colors probably would have been - >sort of a synthetic Foveon situation.
>Has that already been done by the camera in the process of >producing the raw file, or is the raw file really raw and >therefore contains only one color per pixel?
They are usually referred to as photosites on the sensor (some calling them sensels, as a play on sensor+pixels), later interpolated into RBG image pixels.
Each photosite carries a numeric value, no real color. Its number is assigned a color value by its position on the sensor, being under one of 4 R, G, G, B filters in the Bayer array.
> If the latter, >then I assume there's something in the raw file format that >tells the editing software what the color pattern is. And >are there different ways to do the interpolation, so that >you get different results from different software before you >even begin the editing?
During importing the RAW sensor data into the editor it must undergo the transformation to an RBG image in order for you to view it. There are many many RGGB, GBRG, etc. orders of patterns on various sensors. Some even having 4 different colors instead of just 3 colors in a group of 4 (2 green photosites per each red and blue, typical Bayer sensor).
If you would like to test various import interpolation methods to see what they do with the RGGB data converted to RGB data, download Photoline from www.pl32.net and play with it during the 30 day demo period. It has 6 different RAW interpolation methods to choose from, to show you how some are more efficient or accurate than others. (Under: File > Import > Digital Camera RAW Data) Point it at a RAW file to open, then on the left side of the Raw Import preview pane is a drop-down list of various interpolation methods. I most often use the "VNG Interpolation Cubic" option, but depending on the type of fine details in an image, others might be better at times. Zoom in on the image so you can see the individual RGB pixels to see how the interpolation changes with each type selected. If you have a Gretag-Macbeth color chart to photograph, even one found on some web-page and displayed on your monitor screen, that's a good test to see how the interpolations handle various colors and dithering patterns for them. Yellows and oranges seem to be problematic for some RAW interpolation methods. If photographing a Gretag-Macbeth chart displayed on your monitor, photograph it with the lens defocused so the screen's visible pixels don't compound the issue, unless you'd like to see what the interpolation methods also do to fine moire' causing details.
(Before opening any RAW files in Photoline, be sure to go to Edit > Options
> File > Raw, and set the gamma to your workspace, 2.20 if on a PC,
otherwise any RAW images will appear far too dark. Also set the Automatic White-Point to "on" if you are using CHDK RAW files, as they don't contain the white-point information in the RAW header.)
> > Regardless of what it is though, you are correct that it > > must be converted to an RGB format to view it... and > > internally some monitors are necessarily going to > > convert it to a 6-bit format!
>Yeah, I suspect my cheap Toshiba laptop is one of those. >Well, I guess it's just a reminder that you just don't know >what your pictures are going to look like to those seeing >them on a computer.
>Anyway, I just wanted to be clear that you dont actually see >those 14 bits of granularity while you're editing the >picture. You get the benefit of them being there when >making adjustments, but at best you are actually looking >at an 8-bit rendition of the picture.
>In my case it's only 10-bit raw, and I'm just not sure >that's worth the bother.
It depends on your purpose and needs. If you are referring to a CHDK RAW data file (many of them being 10-bit), then you can tweak out slightly more resolution from the RAW file by using more advanced RAW interpolation methods and noise-reduction tools than what the camera uses to create its own JPG file. (Every other year or so someone comes up with a slightly better RAW interpolation mouse-trap.) The increase in resolution by using more advanced methods than the camera uses on the RAW data is slight, but it is observable by photographing resolution test charts or detailed foliage. Those camera are also very good at accurately depicting the full dynamic range of the sensor in the JPG file. But some have found that color information at the extremes of the sensor's dynamic range is more accurately retained in the RAW data. Where a dark red carpet in deep shadows might look nearly black in the JPG file, by using the RAW data it will now appear to have a more reddish hue. In either case, the improvement is slight because of how well those cameras already interpret the RAW data for their JPG images. For some subjects you might find that little bit extra important, so use RAW. You'll have to compare and test it yourself to see if RAW is even worth the bother for your style and requirements of photography. I personally find only about 3 times per month where RAW from my CHDK capable camera is worth the extra steps required for special subjects. The JPGs from them are already really good.
>On Sat, 07 Nov 2009 22:50:12 -0600, Peabody <waybackNO784SPA...@yahoo.com> >wrote:
>>I understand that raw files are typically 12-bit or 14-bit, instead >>of the 8-bit found in jpegs. But when you are editing raw images, >>you don't actually see the 12-bit image do you? I mean, insn't the >>display going to show you an 8-bit picture? So every time you make >>a change to something in the image, the software has to re-render >>the 12-bit info into an 8-bit version for the display.
>>Or do I have this all wrong?
>Correct. (or as someone else already noted, to bit-levels even less than 8) >When printed, reducing this even further.
>12-bit and 14-bit RAW is popularized by "photographers" who must greatly >expand some tonal range in their image during editing, without introducing >large gaps in the recorded colors and intensities. Mostly because they are >nothing but lowly snapshooters dependent on auto-exposure settings, people >that don't know how to correctly expose an image in the first place. Or >because their camera is really poor at doing its own RAW to JPG >conversion--in the process it will clip most of the useful ends of the >dynamic range of the sensor (better, properly programmed cameras do not do >this). Again, proving that they are unskilled photographers because they >don't even know how to purchase decent cameras in the first place.
>This extended bit-range is important for them in editing because:
At this stage the troll is beginning to spout his usual bullshit. Think very hard about the motives he is attributing to people before you accept them.
>(Using exaggerated decimal values of an 8-bit tonal range, for clarity.) >You might have an important area of your image that is highly underexposed >(from poor photography skills). A large and important area of your scene >being recorded in values between 0 and 19 (out of 255 possible 8-bit >levels). In order to make that image the least bit useful the >"photographer" will try to expand that 0 to 19 value range of details to 0 >to 99 in their editor to try to get the details more visible out of those >greatly underexposed areas. You would now only have 20 values representing >almost half (2/5ths) of the image's available hues and intensities. Huge >tonal gaps represented by posterized effects (bands of solid hues) in the >image. Ghastly looking unless used for special artistic effects (known as >posterizing).
>This is why they applaud 12-bit and 14-bit RAW files so much. To help >compensate later for their total lack of skills at photography (wrong >exposures, wrong white-balance settings, ... on ad-infinauseum). Or in not >knowing how to purchase decent cameras that don't require access to the RAW >data, because the full sensor range is already represented in its properly >exposed JPG files.
>Those who preach and worship their God-RAW only reveal their total lack of >skill and talent.
Peabody <waybackNO784SPA...@yahoo.com> wrote: > I understand that raw files are typically 12-bit or 14-bit, instead > of the 8-bit found in jpegs.
The 8-bit JPEG is a (loosely speaking) gamma-encoded value, whereas the RAW data is linear. Our vision is not linear, it's much closer to how JPEG encodes data.
> But when you are editing raw images, > you don't actually see the 12-bit image do you?
You see at least 24-bit images (3 channels at at least 8 bit, maybe more, depends on your system and graphics card), from three fields of sparse 12-bit (or 14-bit) data, namely the red, green and blue photo receptor sites (which are *not* all at the same place).
> I mean, insn't the > display going to show you an 8-bit picture?
Nope. When the conversion tables for calibrated screens come in, it very often has quite more bits already.
> So every time you make > a change to something in the image, the software has to re-render > the 12-bit info into an 8-bit version for the display.
Nope. The software can store the pre-computed image in the image rendering pipeline at the place before you change anything, and just recompute the part afterwards. But just translating 12 to 8 bit, if it would do so, is computationally very cheap and not a problem, unlike many other operations that involve some or many pixels surrounding the one you are actually computing.
> Or do I have this all wrong?
You need to grab a good book or at least a knowledgeable web site and dig in for a week or two, so you'll grasp the basics of digital image manipulation.
> > Well... the RAW data is *never* what you look at > > anyway! It necessarily has to be interpolated to get an > > image format.
> Let me ask about that too. As I understand it, you've got > this Bayer pattern screen built into the sensor, so that > each pixel is really only registering one color. So at some > point, there has to be the interpolation to create for that > pixel what the other two colors probably would have been - > sort of a synthetic Foveon situation.
Correct.
> Has that already been done by the camera in the process of > producing the raw file, or is the raw file really raw and > therefore contains only one color per pixel? If the latter,
Yes, the latter. That's an important reason for shooting RAW.
> then I assume there's something in the raw file format that > tells the editing software what the color pattern is.
No, the pattern for any given RAW format is /usually/ programmed into the RAW software, not the image file. (Not true for DNG though, for example.)
> And > are there different ways to do the interpolation, so that > you get different results from different software before you > even begin the editing?
Absolutely. The de-Bayering algorithm is one of the biggest differences between RAW converters.
> > Regardless of what it is though, you are correct that it > > must be converted to an RGB format to view it... and > > internally some monitors are necessarily going to > > convert it to a 6-bit format!
> Yeah, I suspect my cheap Toshiba laptop is one of those.
You'd be correct.
> Well, I guess it's just a reminder that you just don't know > what your pictures are going to look like to those seeing > them on a computer.
> Anyway, I just wanted to be clear that you dont actually see > those 14 bits of granularity while you're editing the > picture. You get the benefit of them being there when > making adjustments, but at best you are actually looking > at an 8-bit rendition of the picture.
> In my case it's only 10-bit raw, and I'm just not sure > that's worth the bother.
It's a matter of personal preference, a lot of the time. Personally, I shoot RAW on my elderly Canon S30 - which is also only 10 bits- & find it gives me better tonal/colour quality than shooting JPEG.
-- W . | ,. w , "Some people are alive only because \|/ \|/ it is illegal to kill them." Perna condita delenda est ---^----^---------------------------------------------------------------
|
