Re: DCX post comparison of JPEG and CCDRAW from Ni

The images posted at DCX are astonishingly close to each other in quality, but this comparison misses the point, as far as I can tell (given that I don't read Japanese).

The advantage of the RAW format is getting 12-bit image data, which allows you to adjust for color balance, exposure, and contrast without losing latitude in any of the color channels, solarizing images, or getting visible Mach bands. When you make significant changes in the gamma of a single channel (or more) of an 8-bit image, there is usually some loss of detail and latitude. While a loss also occurs in working with a 12-bit image, once the 12-bit image is converted down to 8-bits you can still have a full 256 levels of intensity in each channel (within limits, of course).

The only difference illustrated in the DCX shots is between the camera and Photoshop in capturing the image, converting it to a usable color space, converting to 8-bits, and applying jpeg compression (in camera) or lossless LZW compression (in Photoshop). So you are basically seeing a test of the camera's jpeg compression. While this is interesting and important, it doesn't really show you the great benefits in flexibility to your workflow from using RAW capture (or at least having it available as an option).

Of course, if you've never gotten a bad exposure and never had to use a non-standard light source, this wouldn't be of any interest... ;-)

...Dan

If these photos are intended to show the advantage of CCD RAW versus JPEG, they really missed the boat with me. My eyes may be defective, but the JPEG photos are definitely sharper even though the sky in the second photo is somewhat mottled. Further, the enlargements are way beyond what would be done for a print unless you are delibertly trying produce an unacceptable image.

Rodger

On my trinitron the images were not close in quality...the CCD image was considerably less sharp. I capture CCd Raw images from the Pro-70 and they are as sharp if not sharper than the Jpeg images...It looks like something is wrong with the capture software. ****

Dan Massey said:

The images posted at DCX are astonishingly close to each other in
quality, but this comparison misses the point, as far as I can tell
(given that I don't read Japanese).

The advantage of the RAW format is getting 12-bit image data, which
allows you to adjust for color balance, exposure, and contrast without
losing latitude in any of the color channels, solarizing images, or
getting visible Mach bands. When you make significant changes in the
gamma of a single channel (or more) of an 8-bit image, there is usually
some loss of detail and latitude. While a loss also occurs in working
with a 12-bit image, once the 12-bit image is converted down to 8-bits
you can still have a full 256 levels of intensity in each channel (within
limits, of course).

The only difference illustrated in the DCX shots is between the camera
and Photoshop in capturing the image, converting it to a usable color
space, converting to 8-bits, and applying jpeg compression (in camera) or
lossless LZW compression (in Photoshop). So you are basically seeing a
test of the camera's jpeg compression. While this is interesting and
important, it doesn't really show you the great benefits in flexibility
to your workflow from using RAW capture (or at least having it available
as an option).

Of course, if you've never gotten a bad exposure and never had to use a
non-standard light source, this wouldn't be of any interest... ;-)

...Dan

Click to expand...

I agree guys. The Jpeg was not only sharper but had more detail.

Fevzi

Rodger Carter said:

If these photos are intended to show the advantage of CCD RAW versus
JPEG, they really missed the boat with me. My eyes may be defective, but
the JPEG photos are definitely sharper even though the sky in the second
photo is somewhat mottled. Further, the enlargements are way beyond what
would be done for a print unless you are delibertly trying produce an
unacceptable image.

Rodger

Click to expand...

I agree with your observations about the difference. But that is, I believe, the product of the in-camera software applying a sharpening filter to the RAW image to adjust for the effect of the optical antialiasing filter in the camera.

Of course, since I can't read the Japanese, this is all speculation. And I can't duplicate the test without the camera. There is, however, no way the jpeg is going to be sharper than the RAW file except for extra processing applied in the camera. Maybe they mixed up the samples or, perhaps, the capture software is actually screwed up as you suggest.

...Dan

**** said:

On my trinitron the images were not close in quality...the CCD image was
considerably less sharp. I capture CCd Raw images from the Pro-70 and
they are as sharp if not sharper than the Jpeg images...It looks like
something is wrong with the capture software. ****

Click to expand...

You may be correct Dan...I looked closely at the image and it appeared that there was pixels missing on the CCD and I don't think any sharpening would have improved it...perhaps they were switched...****

Dan Massey said:

Of course, since I can't read the Japanese, this is all speculation. And
I can't duplicate the test without the camera. There is, however, no way
the jpeg is going to be sharper than the RAW file except for extra
processing applied in the camera. Maybe they mixed up the samples or,
perhaps, the capture software is actually screwed up as you suggest.

...Dan

**** said:

On my trinitron the images were not close in quality...the CCD image was
considerably less sharp. I capture CCd Raw images from the Pro-70 and
they are as sharp if not sharper than the Jpeg images...It looks like
something is wrong with the capture software. ****

Click to expand...

Click to expand...

read that as "there were pixels missing" :> )

**** said:

Dan Massey said:

Of course, since I can't read the Japanese, this is all speculation. And
I can't duplicate the test without the camera. There is, however, no way
the jpeg is going to be sharper than the RAW file except for extra
processing applied in the camera. Maybe they mixed up the samples or,
perhaps, the capture software is actually screwed up as you suggest.

...Dan

**** said:

On my trinitron the images were not close in quality...the CCD image was
considerably less sharp. I capture CCd Raw images from the Pro-70 and
they are as sharp if not sharper than the Jpeg images...It looks like
something is wrong with the capture software. ****

Click to expand...

Click to expand...

Click to expand...

Consumer cameras generally measure one color per pixel and then to interpolation to get a color value at each pixel location. Does anyone know what the D1 does? If is does three measurements per pixel, then it is a major step forward in the "consumer" class. If not, then the CCD RAW data is not entirely raw and already reflects significant processing. I would be very interested in hearing if anyone knows what the D1 really does? Regarding the sharpness of the JPEG versions, digital cameras generally generally apply sharpening filters to their output. This would be reflected in JPEG images.

Leon Wittwer said:

Consumer cameras generally measure one color per pixel and then to
interpolation to get a color value at each pixel location. Does anyone
know what the D1 does? If is does three measurements per pixel, then it
is a major step forward in the "consumer" class. If not, then the CCD
RAW data is not entirely raw and already reflects significant processing.
I would be very interested in hearing if anyone knows what the D1 really
does? Regarding the sharpness of the JPEG versions, digital cameras
generally generally apply sharpening filters to their output. This would
be reflected in JPEG images.

Click to expand...

It would stand to reason that the JPEG image had the benefit of whatever post-chip enhancement choices were available in the operating system and that the Raw image simply was what it was on the chip itself. In principle, you can't sharpen a Raw image and still have it as Raw. This could account for the more tweaked looking JPEG.

Not everybody who goes out to shoot an Apples to Apples kind of test has the ability to predict all the variables that need to be flattened before the test reaches a level of "validity". "I'll shoot the JPEG the way I always do..." isn't science-think.

The Raw image is a better starting place for color and sharpness. If somebody teams a $7000 camera/lens combo up with a $99 post-exposure software package, the full potential may be missed, but I'd much rather start with the Raw image for tweaking than the JPEG. It's the digital equivalent of a "negative". I would hope there is a direct path into Photoshop as a 16-bit original.

One thing I noted from the test that was unexpected. Even though the camera had to be touched from shot to shot to change settings, the blow ups show the same detail occupying the same pixel positions. This means that the tripod base (and tripod used) were very solid.

It is the hope of many that the algorithms used in the D1 will filter down to the compact cameras. It is vastly superior to anything previous and produces a much more pixel perfect image.

-iNova

Peter iNova said:

I would hope there is a direct path into
Photoshop as a 16-bit original.

Click to expand...

In this vein, I don't understand the requirement for using Capture through the FireWire interface for getting at the raw images. Nikon's own specifications for the camera cite the number of raw images that can be stored on the flash card. Presumably, then, the images are really on the card. Wouldn't it be possible to write a Photoshop plugin that could deal with them? I realize the plugin would have to deal with color interpolation, defective pixels, etc., etc. What do the plugins for Kodak's raw images do?

Eric

I don't know about Kodak but Canon uses the twain driver and it does little or no processing ofthe image...however PS or Vueprint handles them well...****

Eric Schoen said:

Peter iNova said:

I would hope there is a direct path into
Photoshop as a 16-bit original.

Click to expand...

In this vein, I don't understand the requirement for using Capture
through the FireWire interface for getting at the raw images. Nikon's
own specifications for the camera cite the number of raw images that can
be stored on the flash card. Presumably, then, the images are really on
the card. Wouldn't it be possible to write a Photoshop plugin that could
deal with them? I realize the plugin would have to deal with color
interpolation, defective pixels, etc., etc. What do the plugins for
Kodak's raw images do?

Eric

Click to expand...

**** said:

Eric Schoen said:

Peter iNova said:

I would hope there is a direct path into
Photoshop as a 16-bit original.

Click to expand...

In this vein, I don't understand the requirement for using Capture
through the FireWire interface for getting at the raw images. Nikon's
own specifications for the camera cite the number of raw images that can
be stored on the flash card. Presumably, then, the images are really on
the card. Wouldn't it be possible to write a Photoshop plugin that could
deal with them? I realize the plugin would have to deal with color
interpolation, defective pixels, etc., etc. What do the plugins for
Kodak's raw images do?

Eric

Click to expand...

Click to expand...

The Kodak TWAIN module works with the raw images in a file folder or directly from the camera. The actual raw image is an array of intensity values for each pixel in the CCD. The color channel assignment of the values alternates, RGBG across the CCD, with 50% G pixels and 25% R and B. The intensities are 10 or 12-bit numbers, depending on whether its a DCS 3xx or 5xx/6xx model, representing the image brightness on a linear scale. I do not know how the enpoints of the linear scale are set--probably some combination of the ISO selected on the camera, the exposure, and the compensation--possibly with some analog AGC function in the readout of the CCD as well.

The capture software does a number of things to the raw data.

1. It interpolates the missing RGB values across the entire pixel array, with at least 12-bit accuracy.

2. It converts the linear data to a gamma-weighted output that corresponds to a standard color space Kodak has adopted for their cameras. An optional exposure correction of + or - 2EV can be applied at this time. 12-bit accuracy is maintained in this operation.

3. It balances the color to a standard reference (chosen from a menu) or a sample gray point (chosen from a reference image). Still with 12-bit accuracy.

4. It applies an optional noiise filter to the image (primarily in the B channel). Still with 12-bit accuracy.

I think you will find the D1 does almost exactly the same things, except it gets its color balance information from the 3D matrix exposure sensor and applies an optional unsharp mask to perk up edge contrast lost to the antialiasing filter in the camera optics.

I'm sure I've left out some details, but all this is basically done to acquire the image that is called "raw".

I don't know how Nikon software works, but I see no reason the use of raw images should require the firewire interface (as opposed to just reading the files off the memory card). Nikon has worked with Kodak for many years on the DCS3xx and DCS6xx cameras. I'm sure they have learned their lessons.

The output of the Kodak TWAIN module to Photoshop is, I believe, 16 bits in each of the RGB channels. Of course, the least significant 4 bits are essentially meaningless since they derive from mathematical operations on the 12 bits originally captured from the CCD.

The major difference between all this capture logic on a "pro" camera and a "consumer" camera is whether or not the resulting image is compressed to 8bits intensity depth in each channel and converted to jpeg, both of which take a lot of quality out of the image. Of course, the CCD in the "consumer" camera probably didn't capture the quality in the first place, so it doesn't matter there.

...Dan

Given all you have stated, why do the JPEG examples look better than the RAW examples?

Rodger

I would hazard a guess. The image data off the CCD after color interpolation may be somewhat rough. Camera manufacturers then do various times of processing including but not limited to sharpening to clean their images up and to compensate for idiosyncracies of the CCD. Over the years, I've seen references but few details of the things that are done to images before they are stored. Then they jpeg the images. It should not be surprising to anyone that what comes out of the camera after camera processing is better than what comes off the CCD.

**** said:

**** said:

Dan Massey said:

Of course, since I can't read the Japanese, this is all speculation. And
I can't duplicate the test without the camera. There is, however, no way
the jpeg is going to be sharper than the RAW file except for extra
processing applied in the camera. Maybe they mixed up the samples or,
perhaps, the capture software is actually screwed up as you suggest.

...Dan

**** said:

On my trinitron the images were not close in quality...the CCD image was
considerably less sharp. I capture CCd Raw images from the Pro-70 and
they are as sharp if not sharper than the Jpeg images...It looks like
something is wrong with the capture software. ****

Click to expand...

Click to expand...

Click to expand...

Click to expand...

Sorry but on my Eizo 21'' the ccd raw pics look far superior both printed out and on the monitor. The colours look far more vivid than the jpegs. I would ditch that trinitron if I were you cos you seem to be getting false information.

Michael