JPG sizes from RAW..

[Den]

Does anyone know why shooting in RAW and converting into JPG in Photoshop result in a significantly larger picture and file (5120*3840 (9.1mb!) compared with SHQ JPGs which are 2560*1920 pixels (3.4mb))?

Do you actually get more detail with the larger files or is this just Photoshop interpolating?

[First Light]

Hi Den,

Yes. The answer is that the compression levels are not the same. Remember that you can control the amount of compression used to create a JPEG file. But, unfortunately, it is a "lossy" compression. That means that you lose picture detail when you create a JPEG. The more your compress, the smaller the JPEG file and the more detail you lose. You also begin to introduce weird-looking compression artifacts into the image.

When you output a JPEG from Adobe's Camera Raw import filter in Photoshop, it uses the least compression possible. This creates the biggest file with the least amount of compression.

When you select "Save as..." from the File menu of Photoshop and choose JPEG as your file type, a dialog box will appear that lets you control the amount of compression. You can do this later if you want to create a smaller file (just remember that you'll lose quality when you do this).

You control the amount of JPEG compression in-camera when you select between SHQ, HQ and SQ. But even the highest quality setting (SHQ) compresses the JPEG file more than the highest quality setting of Photoshop. Photoshop simply gives you a wider range of settings for a JPEG file.

By the way, the reason why Olympus also offers TIFF files is because they are uncompressed. It gives you way way to get a totally uncompressed image out of the camera without having to develop a RAW file in your computer. However, RAW files still are the best because they offer a wider dynamic range (12-bits/channel) verses TIFF and JPEG (both 8-bits/channel).

[hammer_400]

FYI studio produces much better jpgs from RAW that are quite alot smaller then in-camera SHQ jpgs... i get around 3-3.5 mb from studio vs 5+ mb SHQ jpgs...

[barondla]

I have seen it stated before that tiff is 8 bit and raw is 12 bit. If so how can they say tiff is uncompressed? You lost 4 bits. That seems lcompressed to me. JUst wondering.
thanks
barondla

[First Light]

barondla wrote:
I have seen it stated before that tiff is 8 bit and raw is 12 bit. If so how can they say tiff is uncompressed? You lost 4 bits. That seems lcompressed to me. JUst wondering.
Normally, when discussing TIFF and JPEG files, "compression" refers to the files, themselves—not the preparation of the raw data beforehand.

But you're right, when the raw data is prepared prior to saving as a TIFF or JPEG file, the net effect is a compression from 12-bits/channel to 8-bits/channel. The number of steps between 0% and 100% for each RGB channel are reduced. But without knowing how Olympus interprets the raw sensor data, it may be that some of the data is simply ignored when a TIFF or JPEG file is created.

However, this does not change the fact that a JPEG file with its "lossy" compression still serves to further reduce the quality of the picture. Therefore a TIFF file will produce a higher qualiy picture than a JPEG. (Bear in mind, these differences are minor—an SHQ JPEG looks very good under most circumstances.)

By the way, TIFF and RAW files can also be compressed. Usually a form of "lossless" compression is used so absolutely no picture quality is lost. For example, you can save a TIFF file in Adobe Photoshop with LZW compression. And some dSLR camera manuacturers compress their RAW files with a lossless compression algorithm so the RAW files are not so big. I wish that Olympus offered this capability for both its TIFF and RAW files.

And there's more: TIFF files are not limited to 8-bits/channel. They can have much greater color depths. For example, Olympus Studio and Adobe Photoshop can each "develop" RAW files into 16-bit/channel TIFF files. You don't actually increase the dynamic range of your picture when you go from a 12-bit/channel RAW file to a 16-bit/channel TIFF but it does give you more headroom in programs like Photoshop for editing the picture without "bruising" the pixels.

[Den]

How useful are the extra bits? I can see how it\'d help minimise damage in post-processing but don\'t most on-line photo printers read everything as 8-bit anyway?

[admin_old]

Here is a Question I don\'t know... If you look in PS, Look at any .ORF you will see Bit depth 16, Resolution 240.
Why does Oly then claim 12 ?

[First Light]

Den wrote:
How useful are the extra bits? I can see how it'd help minimise damage in post-processing but don't most on-line photo printers read everything as 8-bit anyway?
Their usefulness depends on the picture. For many photos, you won't miss them. It's the difficult or problem photos where the extra data bits can be a shot saver. For example:

Overexposed and underexposed shots can be corrected by adjusting the Exposure Compensation setting when you develop a RAW file. In some cases you can recover "blown" parts of an overexposed shot. And you can recover more shadow detail from an underexposed shot.
More highlight and shadow detail can be preserved in shots with a wide dynamic range. For example, you shoot a scene with both shadows and bright sunlight. Using the RAW file you maintain more image data as you "develop" into a 16-bit/channel color space in Photoshop. Then you have more headroom to adjust the image for printing. You can use Photoshop's Shadow/Highlight adjustment controls to manually compress the dynamic range the way you want. Plus you now have control over which, if any, parts of the image you want to clip if you later downsample to an 8-bit/channel space.
You can employ tricks of the trade like "exposing to the right" (of a histogram) in order to optimize the dynamic range of your camera to the scene and then adjust the exposure back as desired when you develop the RAW file.
If you plan to heavily edit a shot, having more data bits and developing into a 16-bit/channel space will enable you to reduce pixel "bruising". For example, if you darken part of your photo in an 8-bit/channel space you will often lose some image detail. A pixel at 1% may be pushed to 0% and you can never get the 1% value back. With 16-bits/channel you have more levels and less clipping of detail will occur. Similarly, if you adjust the "levels" of your image in an 8-bit/channel space, you will reduce the number of level steps in part of your image. In a 16-bit/channel space you more steps and can minimize the loss.And there are other advantages that you get with RAW files that do not relate to more data bits. You can adjust the white balance (WB ). You have the best possible copy of your image to archive for future use when development tools are more advanced. Et cetera.

[First Light]

tspore wrote:
Here is a Question I don't know... If you look in PS, Look at any .ORF you will see Bit depth 16, Resolution 240.
Why does Oly then claim 12 ?
The reason Photoshop "sees" a 16-bit/channel depth is probably because it doesn't support anything in between 8-bits/channel and 16-bits/channel. When it sees that the depth is greater than 8-bits, it probably just moves to the next higher setting. Plus I don't think that many of the file formats that support channel bit depths higher than 8 can do in between depths. (If they do, then the files may not be useable by most programs.) These file formats (like PSD and TIFF) probably require a jump to 16-bits/channel also.

I have no idea about the resolution. Maybe Olympus encodes that information in a way that Photoshop can't read so it just assigns a value of 240.

[kengreen]

I may be mistaken, but I beleive you can change the settings in ACR both regarding the resolution and the bit depth. If I remember the four extra bits are empty and do not affect image.
Ken

[source]

Den wrote:
How useful are the extra bits? I can see how it\'d help minimise damage in post-processing but don\'t most on-line photo printers read everything as 8-bit anyway?

One of the best explanations I\'ve come across about this, that also explains why and how to expose to the right can be found at Luminous Landscape (http://www.luminous-landscape.com/tutorials/expose-right.shtml)... although you might all know this article already.

[admin_old]

That is a great article. I don\'t think that I have seen it before.
Thanks.
ken in PS you can change 8 Bi 16 Bit 32 Bit. But someone said the the .ORF\'s are 12 Bit depth, if I am not mistaken, It just seems like an odd number. So I was trying to figure out why.
Thanks
Tony

[First Light]

tspore wrote:
...But someone said the the .ORF's are 12 Bit depth, if I am not mistaken, It just seems like an odd number. So I was trying to figure out why.
Short Answer:
First, page 173 of the E-1 Reference Manual states that Olympus raw data contains 12 bits (per channel).

Second, every manufacturer of images sensors wants to get the best signal-to-noise ratio so they can detect the most possible levels between black (no light) and white (maximum light). 12-bits/channel is simply the best that Kodak has been able to achieve (so far) in a production 4/3rds sensor. Fuji was able to create a larger 14-bit/channel sensor by using a combination of large and small pixels.

Technical Answer:
The Kodak 4/3-type sensors used by Olympus so far, along with their associated support circuitry, are able to detect about 4096 levels of luminosity. This corresponds to 2^12 (2 raised to the 12th power) levels of luminosity or 12 data bits. Note: A single data bit has two states—on and off—and this is respresented by the number "2". The support circuitry may adjust the data signals based on the temperature of the sensor (most image sensors are sensitive to heat) and may provide baseline noise reduction. I believe that the 12-bit/channel raw data is the output after the support circuitry has adjusted it.

The image sensor, itself, is color blind. You can think of it as a grayscale sensor. Color filters are placed in front of the sensor receptor sites to generate color data. Plus an IR-blocking filter is used to reduce the sensitivity to infrared light (most image sensors are very sensitive to IR). The pattern of the receptors and the type and distribution of color filters is as much art as science. Olympus uses red, green and blue (RGB ) color filters. Other manufacturers, like Sony, have added a fourth color filter to expand the color gamut. The raw data does not translate directly to RGB values. Rather, sophisticated algorithms are used to translate the sensor data into RGB channels per pixel.

[source]

tspore wrote:
The image sensor, itself, is color blind. You can think of it as a grayscale sensor. Color filters are placed in front of the sensor receptor sites to generate color data. (...) The pattern of the receptors and the type and distribution of color filters is as much art as science. Olympus uses red, green and blue (RGB ) color filters.

It is really remarkable how conceptually similar is the surface of one of these chips to the surface of our own retinas. Chips have red, green and blue sensing elements (by virtue of the filters on top of them), and our retinas have three kind of cones (each sensitive to red, green and blue light), as well as rods as you all know. And it is the neurons at the deeper layers of the retina, and ultimately our brain, the ones that group different adjacent cones together, to form \"a pixel\" and make sense of it all, much like the algorithms inside our cameras or Adobe Camera Raw.

Post edited by: source, at: 2005/10/11 22:48

[admin_old]

There is a link in the \"link\" section which has info about the Kodak sensor.
links (http://fourthirdsphoto.com/component/option,com_weblinks/catid,18/Itemid,36/)

[russ.will]

As per usual a simple question has thrown up a large amount of useful information. However Den\'s original question was not about image compression and bit depths. He wants to know why his images pixel dimensions have been doubled when converting from a .ORF to jpeg in ACR. I\'ve not noticed this phenomenon, but I have seen this question asked before (possibly on the old site) and would be interested to know, compression ratios and bit deths not withstanding.

Russell

Post edited by: russ.will, at: 2005/10/12 10:49

[admin_old]

ops... My bad.
Here it is as I understand it.
All jpeg\'s have compression.
In PS 12- is very little 3 - a Lot .
If you go from 12 to 10 in jpeg quatily the file is much smaller.
I usually save at 9 or 10.
I don\'t notice any difference.
T

[First Light]

Hi Russell and Den,

For some strange reason I completely overlooked the part of the question about the raw (orf) pictures having more pixels after "development". :( Sorry about that. The answer is fairly simple:

At the bottom left corner of the ACR window is a "Show Workflow Options" checkbox. Check that box to display the options if they are not already displayed. You'll see a "Size" option where you can select the size that you want your raw (orf) pictures to be converted to when they are "developed" into Photoshop.

Evidently, this size option is set larger than the original size of your pictures as shot by your camera. Simply change this setting to match the original pixel dimensions of your image and the problem of unexpected enlargement should go away.

By the way, I would never resize an image at this stage in the workflow. It would be better to wait until the image has been developed into Photoshop and then use the "Image Size" dialog to resize the picture. It gives you much more control, including the ability to select the resampling algorithm (bicubic, etc.).

[russ.will]

I see. I\'ve never bothered altering this option so it\'s never been a problem. I presume it defaults to the last size set on the next image, hence Den\'s huge jpegs.

I always resize images prior to sharpening. I tend to use Bicubic smoother for larger and bicubic sharper for smaller. Is this right?

Russell

[First Light]

russ.will wrote:
...I always resize images prior to sharpening. I tend to use Bicubic smoother for larger and bicubic sharper for smaller. Is this right?
Sharpening – Yes, sharpening should be the last step in most workflows.

Resampling algorithms – Adobe originally intended the Bicubic Sharper algorithm for size reduction and the Bicubic Smoother algorithm for size enlargement. However, some pros have found that the Bicubic Sharper seems to work better for some enlargements, too. Which works best for you may vary from image to image, depending on the contrast and level of detail. Don't be afraid to experiment. Use adjustment layers so you can swiftly compare one method to another and select the one that works best. Then record actions of your favorite procedures so you can easily repeat them later in a single step.

Fred Miranda recommends that enlargements be made in small multiple steps for best results. For some good tips see Chapter 4 of "The Photoshop CS2 Book for Digital Photographers" by Scott Kelby. I can't recommend this book highly enough—it's my favorite and Scott has a great off-the-wall humor.

[carman]

Does anyone know why shooting in RAW and converting into JPG in Photoshop result in a significantly larger picture and file (5120*3840 (9.1mb!) compared with SHQ JPGs which are 2560*1920 pixels (3.4mb))?

Do you actually get more detail with the larger files or is this just Photoshop interpolating?

The increase in size is due to PS interpolating the image. You should reset to the default. This will give no increase in size. PS cannot put detail in that was not there in the first place. This is only of use for large prints. Raw conversion does not automatically result in a larger file.