PDA

View Full Version : The wonderful and missunderstood world of JPEG



Puffer Fish
11-24-2007, 18:38
There are two issues on the digital world that people, in general, have difficulty understand... white balance and the issue of RAW versus JPEG.

Before getting to the examples, both good and bad... look at this review of the Fuji DSLR...

Fujifilm S5 Pro Review: 16. Software: Digital Photography Review (http://www.dpreview.com/reviews/fujifilms5pro/page16.asp)

No difference between the JPEG and the RAW??? How can that be?

Puffer Fish
11-24-2007, 18:55
First off, a camera JPEG is more than just a compression of the image, the camera also does a lot of things to the data...it applies some sort of sharpening. It makes some contrast and color adjustments. It applies the white balance calculation....here, we are just going to look at the JPEG... those other things do a lot...but lets start with one part at a time.

As a reminder, a picture is just a bunch of "dots", that each one has some numbers attached to. Put all the dots together, and you will have an image. For the most part, if you are using windows, that is something called sRGB, where the R= Red, G=Green and B = Blue. three numbers for each dot, and you have a color... all the color you can see on a screen are represented by 0 -256 for each value. That equals over 16,000,000 possible colors, which is more than any human can see.

Puffer Fish
11-24-2007, 19:02
Second, JPEG is not a single standard, it is the format for a whole bunch of standards. There are (depending on how one looks at the), either 12 steps, with 12 types of chroma subsampling.... for 144, or 99 (1 to 99%) times 12 for 1188 different versions, add in some other forms, and one gets a confusing number.

Interestingly, almost every camera maker uses a different type, some that work great, some that don't. And it turns out, there is a way to know which they are using...

Puffer Fish
11-24-2007, 19:19
The resolution chart, on DPreview.com, is the same image each time, so the resulting JPEG is a reflection on the compression method they use... here are some common camera's

Note: I have adjusted the number to a 10 meg size, so the amount of compression is the same:

Canon EOS 40D -------- 2645
Nikon D200 -------------2728
Pentax k10D-------------2271
Nikon Coolpix 5100 -----2732
Canon Power shot G7----3330
Fuji F50fd ---------------3636
Canon Power shot G9 ---3432

Notice anything odd.... why would the smallest camera, have the largest JPEG.. and why would the DSLR's, have the most compression?

This might help... the Fuji S5----4346

Turns out, a lot of DSLR's don't expect one to use the JPEG, so they put a not so good JPEG in their camera.

Fuji, actually expects their JPEG to be used, and it is a lossless JPEG...


Note: A bit of the size difference is also due to the amount of detail, the more there is in the image, the larger the file will be... this does not represent that much, and one could easily point out the DSLR's actually do have more data, so the difference is actually bigger than shown.

Puffer Fish
11-24-2007, 19:21
From the above, one would highly recomment, that with most DSLR's you only use RAW, but that is not the case with the Fuji. If you own a camera, you should check this, to see what they are using.

Cheddarchick
11-24-2007, 19:21
Thanks Puffer....I love reading about photography

Soonerwink
11-24-2007, 19:24
Thanks Puffer keep the threads coming, I am learning alot from them.

Puffer Fish
11-24-2007, 19:35
Ok, so we know they are using different ones.. but which ones, and does it effect the picture?

One would assume that compression is bad... but not in this case...

Remember there are degrees of compression, and different amounts of chroma subsampling.

I will skip over the details of what the JPEG program is really looking at (post if you would like the technical details), the the first level is called:

YCbCr 1x1 1x1 1x1

This is a completely lossless pattern, what it saves will be exactly like the original, if the compression is set to 1%.

I used an image from DPreview, that is a raw D200 picture he uses for testing. I checked the detail (the exact distribution of the pixels and their individual counts) and saved this image with a 1x1 1x1 1x1 and 1%... here is a small section to compare the information:

http://i78.photobucket.com/albums/j98/garinj/d200masterimage.jpg

Here is the compressed JPEG:

http://i78.photobucket.com/albums/j98/garinj/d200jpegbestdetail.jpg

If you have any good photo software, please download and compare these two images.

Puffer Fish
11-24-2007, 19:36
Thanks Puffer....I love reading about photography
Thanks... am trying to provide as much hard information as possible.

Puffer Fish
11-24-2007, 19:37
Thanks Puffer keep the threads coming, I am learning alot from them.
Thanks for taking the time to read them...

Puffer Fish
11-24-2007, 19:39
Those two images are not just close, but identical, pixel by pixel... and this represents a "compression" down to 4.8 meg... from the original 10 meg.

But that is a lot bigger than Nikon, for example uses. So what happens if you go with more compression?

Puffer Fish
11-24-2007, 19:42
Next, I took the same chroma, and increased the compression up to 25% -

http://i78.photobucket.com/albums/j98/garinj/d200masterimage.jpg

This is the compressed one:

http://i78.photobucket.com/albums/j98/garinj/d200jpegnone25.jpg

The image size was down to a small 1.08 meg... and I can easily tell something has happened to it, but I don't know if someone looking could tell.. again, download and look for yourself.

Puffer Fish
11-24-2007, 19:45
Here is what happens if you keep going, down to 50% compression and a size of .4792 meg...

Again the original versus the compressed one:

http://i78.photobucket.com/albums/j98/garinj/d200masterimage.jpg


And Compressed:

http://i78.photobucket.com/albums/j98/garinj/d200jpegnone50.jpg


This has some many JPEG artifacts, that I would never use it... how does it look to you?

Puffer Fish
11-24-2007, 19:49
The large picture, in the above example, has the sky with bands of light to dark... and as you can see, if you look closely, the image is getting blurry.. this is the problem with the DSLR JPEG's and why you need to use RAW with them.

RoadRacer1978
11-24-2007, 19:53
Most of the comparisons were pretty close but the last one you could easily tell a big difference. Everything got really pixelated. I know from using photoshop and doing some graphic editing it is much easier to work with a high resolution image. You can alway compress it and reduce the size once you have finished working on the image, but you can't put in resolution that wasn't there to begin with. Not sure that this ties in to your posts, but something I have experienced from the enhancement phase of taking photos.

Puffer Fish
11-24-2007, 20:00
Before going on, the information presented here has been staring all of the experts in the face for some time, just that no one has put the pieces together.. everyone has been to busy looking at the results, and not been spending time looking at why they are getting the results.

Ok, off the soap box now.

Puffer Fish
11-24-2007, 20:01
Most of the comparisons were pretty close but the last one you could easily tell a big difference. Everything got really pixelated. I know from using photoshop and doing some graphic editing it is much easier to work with a high resolution image. You can alway compress it and reduce the size once you have finished working on the image, but you can't put in resolution that wasn't there to begin with. Not sure that this ties in to your posts, but something I have experienced from the enhancement phase of taking photos.
It will in just a minute...hold on please.

Puffer Fish
11-24-2007, 20:04
I had pointed out earlier, that there were twelve differ Chroma subsampling methods... each will reduce the number of variables by a different factor...

However, only 1x1 1x1 1x1 and 2x1 1x1 1x1 are commonly used, as reducing the others is easy for a person to see. Ideally, everyone is using the 1x1 1x1 1x1, but who know? Or is there a way to tell?

Puffer Fish
11-24-2007, 20:13
Well, there is a way... all you need is a picture taken with the camera, and a program that can do the jpeg compression. You take the image and compress it until it is nearly identical to the original compression, and you have it (there will be a slight difference, due to rounding of numbers and the exact math).

So, back to the Nikon D200 image... what did they use? Turns out to be 6% (it shows a 4% compression, but 4 % does not yield the size they have) compression. Can you see 6%.. not very well... only in the finest of detail could you tell. Guess what? That is what they are seeing when they look at raw versus JPEG from that camera. No magic, nothing fancy... just what you should see.

Puffer Fish
11-24-2007, 20:56
What about the Fuji f50? First off, that 6% is a very common compression... the F11 I have, uses the same amount. But that is not the case with the f50, it uses (again, it shows 3.2), but measures something around 2 (there are a lot more subtle adjustments that each company can make, so the information here is just approximate. The type of scanning it does also effects the size a bit...but has no effect on quality.

Does this loose some information - yes it does, but only a very, very small amount, and looking at the detail, there is some damage, but no loss of information.. resolution appears to be intact.

Here is an original tiny section:

http://i78.photobucket.com/albums/j98/garinj/fujimasterdetail.jpg

A two percent compression:

http://i78.photobucket.com/albums/j98/garinj/fuji50jpeg2.jpg

And for comparison, a 6% compression"

http://i78.photobucket.com/albums/j98/garinj/fuji50jpeg6.jpg

You will have to blow this up a great deal to see the difference... but it is there...

Puffer Fish
11-24-2007, 20:59
So, bottom line, know what your camera is doing with compression... and let that be your guide... With the fuji, there, for all practical purposes, would be no advantage having raw... with most DSLR's, you need raw to get all of the detail... But coming up next is the effect of other "adjustments" the camera might make and how they effect the image quality.

bversteegh
11-27-2007, 00:09
You are missing one of the major benefits of RAW processing - bit depth. Now if all your exposures are perfect (and strobe positioning, so you don't have shadows); JPEG is great, and eliminating the time associated with RAW processing is a real time saver.

But, if you missed the exposure, white balance, or need to pull out detail in shadows, RAW has a huge advantage. RAW uses the full bit depth of the image (12 bits or even 14 bits on the new Canon DIGIC III's) - so you have tons more dynamic range to play with. This allows you to adjust the image to optimize the histogram, and keep continuous tones.

Remember how digital numbers work - each additional bit doubles the amount of potential information available (but also increases the file size). An 8 bit image has 256 discrete levels per channel, 12 bit images have 4096 discrete levels for each color channel (typically RGB). So if you have an underexposed image (or want to see details in shadows) - working in RAW allows you to extract more levels without losing detail when you adjust the image. So after adjusting an image, you can still have continuous tones, which look more pleasing to your eye.

So why don't RAW images always look better if they have more information? Your computer display only uses 8 bits/color (times the 3 color channels, so you can see millions of colors) - so if your original JPEG takes advantage of the full dynamic range of the display, it will display equally as well as the RAW image (and the camera will have specific tone curves and sharpening built in to the processor - saving you the time to try and optimize your image).

So is RAW always better? No, but it is more flexible when you try and "save" an image with good focus and composition, but lighting/color problems.

Puffer Fish
11-27-2007, 06:05
You are missing one of the major benefits of RAW processing - bit depth. Now if all your exposures are perfect (and strobe positioning, so you don't have shadows); JPEG is great, and eliminating the time associated with RAW processing is a real time saver.

But, if you missed the exposure, white balance, or need to pull out detail in shadows, RAW has a huge advantage. RAW uses the full bit depth of the image (12 bits or even 14 bits on the new Canon DIGIC III's) - so you have tons more dynamic range to play with. This allows you to adjust the image to optimize the histogram, and keep continuous tones.

Remember how digital numbers work - each additional bit doubles the amount of potential information available (but also increases the file size). An 8 bit image has 256 discrete levels per channel, 12 bit images have 4096 discrete levels for each color channel (typically RGB). So if you have an underexposed image (or want to see details in shadows) - working in RAW allows you to extract more levels without losing detail when you adjust the image. So after adjusting an image, you can still have continuous tones, which look more pleasing to your eye.

So why don't RAW images always look better if they have more information? Your computer display only uses 8 bits/color (times the 3 color channels, so you can see millions of colors) - so if your original JPEG takes advantage of the full dynamic range of the display, it will display equally as well as the RAW image (and the camera will have specific tone curves and sharpening built in to the processor - saving you the time to try and optimize your image).

So is RAW always better? No, but it is more flexible when you try and "save" an image with good focus and composition, but lighting/color problems.

Thank you... which is back to the major difference between a point and shoot and a DSLR... it is much harder to get a perfect exposure with a P+S...and most images cannot be "fixed" to any degree.

Why is it only one in a few P+S images are really good... that is one major part of it.

But the human part also comes into play... most people that invest the money in a good DSLR also spend the time to learn more about it. Most people that invest in a P+S don't. As the more forgiving is the DSLR, that drives the image quality of the Point and Shoot down even farther. Great images from a P+S actually take more knowledge and understanding...not less.

And working in 14 bit is one serious upgrade from 8 bit.

Puffer Fish
11-27-2007, 17:20
I did some checking on the concept of Bit depth... with some camera's it has a bit of validity. Particularly the greater dynamic range that the Fuji S5 has. However, that is a professional camera, designed for taking pictures of people in a studio. It has terrible resolution, and great dynamic range. Your typical high end consumer DSLR is not so lucky.

There is no doubt that the more bits the smaller the adjustment can be made... and that finer resolution prevents image degradation. But in the end, it is translated into a JPEG to be seen...so whatever those changes might be.. they are long since gone.

Even the concept of adjusting an image in 12 bits, while viewing it in sRGB or the far better apple standard, leaves the whole issue in doubt. As we cannot see the 16,000,000 colors in RGB, the 100,000,000's of million more represented by a lot more bits is, at least at this point, rather meaningless.

The best system (I hate sRGB by the way) is one that starts in the same format and stays there thru the whole process. Adobe, does not do that, and suffers translation errors every time you open an image and do anything with it..Adobe Raw is not better.

Note: for those that don't know, there are several ways to display a color... the JPEG one is different from RGB, which is different from the internal method used by Adobe..which is a Lab system. JPEG is usually a YCbCr system..raw uses a one of several systems.

Note: some JPEG's actually use RGB, but not most.

So to do a print, one usually is going from camera raw... to screen RGB, to (if using Adobe) internal Lab, then to YCbCr and if printed to CYMK. Any micro detail is completely lost in the first step.

A clean system involves using the same standard all the way thru the system can produces amazely accurate color. Sadly, sRGB is perhaps the weakest link in the process.

bversteegh
11-27-2007, 23:47
I did some checking on the concept of Bit depth... with some camera's it has a bit of validity. Particularly the greater dynamic range that the Fuji S5 has. However, that is a professional camera, designed for taking pictures of people in a studio. It has terrible resolution, and great dynamic range. Your typical high end consumer DSLR is not so lucky.

There is no doubt that the more bits the smaller the adjustment can be made... and that finer resolution prevents image degradation. But in the end, it is translated into a JPEG to be seen...so whatever those changes might be.. they are long since gone.

Even the concept of adjusting an image in 12 bits, while viewing it in sRGB or the far better apple standard, leaves the whole issue in doubt. As we cannot see the 16,000,000 colors in RGB, the 100,000,000's of million more represented by a lot more bits is, at least at this point, rather meaningless.

The best system (I hate sRGB by the way) is one that starts in the same format and stays there thru the whole process. Adobe, does not do that, and suffers translation errors every time you open an image and do anything with it..Adobe Raw is not better.

Note: for those that don't know, there are several ways to display a color... the JPEG one is different from RGB, which is different from the internal method used by Adobe..which is a Lab system. JPEG is usually a YCbCr system..raw uses a one of several systems.

Note: some JPEG's actually use RGB, but not most.

So to do a print, one usually is going from camera raw... to screen RGB, to (if using Adobe) internal Lab, then to YCbCr and if printed to CYMK. Any micro detail is completely lost in the first step.

A clean system involves using the same standard all the way thru the system can produces amazely accurate color. Sadly, sRGB is perhaps the weakest link in the process.

You absolutely missed the point and the benefit - it isn't related to color space at all - it is all about how the dynamic range of the image is used to preserve as much information as possible in the image. And almost any camera that offers RAW has at least 12 bits native dynamic range - the JPEG processing in the camera uses algorithms tuned for that camera to resample the image to 8 bits/channel when it saves the file.

But the point is this - let's assume you missed the exposure by one stop, you have lost a whole bit of dynamic range. With your 8 bit Jpeg, you now have only 128 levels of good information (and it is gone, you can't get it back). When you adjust the exposure in your image to use the full dynamic range of the display (256 levels; full black to true white), you don't have smooth transition between tones, which will be evident in your image. As an oversimplification (and admittedly this is not a fair comparison) - leave your monitor at the same resolution with a picture displayed, and switch to lower color resolutions (go clear down to 256 colors if your video card supports that mode) - and you will find the image much less pleasing, because you have lost the continuous tone transitions - but you still have exactly the same resolution/sharpness of the image. This is an extreme example of what happens when you lose dynamic range from the image.

Do the same thing in 12 bit space - the one stop exposure costs you 1 bit dynamic range again - and the unadjusted picture is still dark and underexposed. But you have 11 bits of dynamic range left. adjust the image using histogram or levels tool, and correct the exposure so you now have highlights and whites in your image. When the raw software converts the corrected image with 11 bits image to 8 bits for display when you convert to a JPEG, it resamples and compresses the oversampled image- you end up with an image that still has continuous tones. This is the whole basis of digital sampling theory.

This same advantage allows you to bring out detail in shadows - again, you can use the levels tool to selectively expand the dynamic range in the shadows - and since you started with a 12 bit image, you have some information available. Save it as a JPEG, and you will have detail that would otherwise be lost.

As I said in the first post, shoot a properly exposed and lighted picture, and RAW offers no advantage - and it is a whole lot less work to get your image.

But those extra bits have real value when you are fixing mistakes. This is not Adobe propoganda - it is a very real benefit with real world application.

Puffer Fish
11-28-2007, 07:47
Color space is, by definition, the representation of what any color measuring system can "see". If you have a larger dynamic range, then you have, by definition a larger color space.

To say that because you have 12 bits or 14 bits, inherently means you have a bigger dynamic range, is not necessarily true.

Note: You really should use a raw image with most DSLR's, and you should because of increased dynamic range, but the issue is the loss of that range because of JPEG compression that the camera uses. Not because of the number of bits.

An example:

I have two spectrophotometers here.. one cost $800 and reads in tenths, one costs around $28,000 and reads in .00. So my high end one has ten times the dynamic range? Nope... they both have exactly the same color space, just that one cuts up the space ten times smaller.

As we humans, at best, in some areas of the spectrum, can see down to around .3 on the scale they use... even the .1 is more accurate than we can tell.

Why buy the $28,000 one? Well it has to do with calibration, repeatability across machines and long term stability... not color space.

If you want to see this for real... take your under exposed raw image... and load it into paintshop pro (not Adobe) and save the image as a lossless jpeg.. (1% 1x1 1x1 1x1).. then adjust each of them, and check the results... they come out the same. Even when the raw is corrected in adobe... the compared results don't show any difference.

Note: you can try doing all of this in Adobe, but there is one more conversion that takes place and that results in some compression also...try it you will see.

When I use the word "conversion", what I am talking about is the conversion from one color space to another. Most of the time, neither you or I know how that was done (although you can see and measure the effects), but in principle, you have three main ways:

1. Direct mapping from one to another. This is what you are assuming is happening and with the bigger color space, to a smaller space, you will loose the edges.

2. Relative mapping from one to another. This puts the new point, in the same relative space, and results in color shifts, but not data loss. People don't usually like that, but it is effective.

3. Edge proportional, where the middle is mapped the same, and as you approach the edge, it is made to fit, so that you don't loose data, but you now have distorted image data. This looks better than two, but one cannot got back and forth with accuracy.

In Adobe's case, they are comparing their raw against the camera's JPEG.. which sucks for totally different reasons. And their own JPEG system causes a whole new set of issues.

On that issue, really accurate color management is not possible using Adobe, as the translation errors are so large. Just try saving a copy of a copy of a copy of a copy in adobe and watch what happens to the image. If you use the Fuji or Nikon professional software, that does not happen, but they don't do translations. Fuji, in particular, uses a RGB JPEG... and processes in RGB. Not the best color space, but they don't have translation compression.

Scubakraut
11-28-2007, 20:15
Thanks for the great guide! How does the Canon Digital photo professional process the image? I know I can convert it to jpg later if I like.

Puffer Fish
11-28-2007, 23:47
Thanks for the great guide! How does the Canon Digital photo professional process the image? I know I can convert it to jpg later if I like.
I hate to say this, but I don't know... will see if I can find out tomorrow....

Scubakraut
12-04-2007, 14:41
The resolution chart, on DPreview.com, is the same image each time, so the resulting JPEG is a reflection on the compression method they use... here are some common camera's

Note: I have adjusted the number to a 10 meg size, so the amount of compression is the same:

Canon EOS 40D -------- 2645
Nikon D200 -------------2728
Pentax k10D-------------2271
Nikon Coolpix 5100 -----2732
Canon Power shot G7----3330
Fuji F50fd ---------------3636
Canon Power shot G9 ---3432

Notice anything odd.... why would the smallest camera, have the largest JPEG.. and why would the DSLR's, have the most compression?

This might help... the Fuji S5----4346

Turns out, a lot of DSLR's don't expect one to use the JPEG, so they put a not so good JPEG in their camera.

Fuji, actually expects their JPEG to be used, and it is a lossless JPEG...


Note: A bit of the size difference is also due to the amount of detail, the more there is in the image, the larger the file will be... this does not represent that much, and one could easily point out the DSLR's actually do have more data, so the difference is actually bigger than shown.

The Fuji S5 picture quality is worse in my eyes then the Canon 40D or Nikon D200.
Somehow if I do my own conversion supplied with the Canon SW, the jpg quality comes out worse then if I took the same shot using the cameras supplied internal jpg compression. Any explanations for this?

Puffer Fish
01-07-2008, 10:10
Just noticed your post.. sorry for not responding.

Image quality... how sharp and how much detail is not the first thing that happens with compression of color space. Black and white edges will be preserved far more than subtle color variations in the sky, for example.

Eventually it does, but not in the first stages.

Think of this (which is close, but not exactly what happens) as the JPEG size reflects the number of different colors shown, and the bigger the number, the larger the color space is (or the more detailed the color space is). Color detail may or may not reflect how sharp the images look.

Fuji does their's by using one big pixel, and one small pixel, mixed together. The small ones for bright lighting and the big ones for low light. This gives a bigger range than any one size sensor could do, but the small sensor has empty space around it... and cannot see as well in low light, so the end result is a loss of resolution (what you can see in the detail).

So you get a bigger color space, and more detailed color within the space, but you loose resolution.

Is it worth it? Super detail is not needed in portrait taking, but color accuracy is (at least in Japan), so yes, it is.

Would I buy that camera for uw pictures? No... I would not.

Just as you should use raw from any DSLR, but for different reasons than most people think.

Regarding the canon software... it is not designed as the commercial Nikon and Fuji versions. Please keep in mind that your "JPEG" is not the same as another camera's is....how much compression is done is not something that is obvious to the user. The high end Fuji comes with the software, but that software is not provided with consumer cameras.

The resulting image size from any JPEG processing, is an excellent measure of the amount and type of compression... if you know the size, you can figure out how much compression is goin on...but you need to have the same reference image to compare.. which is why the DPriew site is handy. In the example I posted, completely loss less is about 200 k larger than the S5 jpg image...



The resolution chart, on DPreview.com, is the same image each time, so the resulting JPEG is a reflection on the compression method they use... here are some common camera's

Note: I have adjusted the number to a 10 meg size, so the amount of compression is the same:

Canon EOS 40D -------- 2645
Nikon D200 -------------2728
Pentax k10D-------------2271
Nikon Coolpix 5100 -----2732
Canon Power shot G7----3330
Fuji F50fd ---------------3636
Canon Power shot G9 ---3432

Notice anything odd.... why would the smallest camera, have the largest JPEG.. and why would the DSLR's, have the most compression?

This might help... the Fuji S5----4346

Turns out, a lot of DSLR's don't expect one to use the JPEG, so they put a not so good JPEG in their camera.

Fuji, actually expects their JPEG to be used, and it is a lossless JPEG...


Note: A bit of the size difference is also due to the amount of detail, the more there is in the image, the larger the file will be... this does not represent that much, and one could easily point out the DSLR's actually do have more data, so the difference is actually bigger than shown.

The Fuji S5 picture quality is worse in my eyes then the Canon 40D or Nikon D200.
Somehow if I do my own conversion supplied with the Canon SW, the jpg quality comes out worse then if I took the same shot using the cameras supplied internal jpg compression. Any explanations for this?

Scubakraut
01-07-2008, 10:37
ok, thanks for the insight. I do like taking my photos in Raw, but once I have my sharpness and white balance set, I'd like to store my end results in jpg format about half the size what the original raw is. Otherwise I soon have no space to store all my nice vacation photos. One answer could be get a bigger external Hard drive.
Do you keep all your raw's just to view them as slide shows later?

Puffer Fish
01-07-2008, 14:41
ok, thanks for the insight. I do like taking my photos in Raw, but once I have my sharpness and white balance set, I'd like to store my end results in jpg format about half the size what the original raw is. Otherwise I soon have no space to store all my nice vacation photos. One answer could be get a bigger external Hard drive.
Do you keep all your raw's just to view them as slide shows later?
Save on a DVD (be careful on the format... as I have lost one due to some strange reading issue) and make a copy. Then save in a JPEG with zero compression. Leave that on your computer. If you work on the image, do it on the JPEG, but save as another file. Even zero compression will save a ton of space. A 10 meg image, in JPEG, with no compression is typically around 4 meg (give or take a meg), and you know that a raw is way larger than 10 meg.

bversteegh
01-27-2008, 01:39
DO BOTH - get an external hard drive (if you watch, you can get 350 gb drives for $100); plus burn DVD's.

I actually set up an old computer as a raid server - so all my original stuff has redundant backups. I had a good friend loose an entire dive trip when he couldn't read the DVD's (the computer he burned them on died, and they wouldn't read on anything else).

Over the years, I have had 2 total disk crashes (one laptop, one desktop) - the backups saved me.