[robski]

Originally Posted by Stephen

A jpeg file does not necessarilly contain the same detail, as we know up to 75% of the pixels have been chucked away, only to be put back as a best guess when reopened.

I see we are sliding back into the realms of rhetoric. This statement maybe fair if the jpeg encoder is set with co-efficients to give extreme compression but in our case we are talking about using the best quality setting available. (it is possible to have finer setting but this would not give any worth while compression and lossless techniques would be better) For my sins I have had to get my head around the internal workings of the jpeg encoder in my line of work.

My interest in this thread had been to gain enough insight to see whether it is worth investing in more hardware to use RAW effectively. Does it give me enough latitude to correct problem shots. Many have talked about the improved sharpness which has not been my experience. I can see from some examples in this thread there is a difference but I believe some models of camera seem to blur the image before passing to jpeg encoder. This approach would achieve smaller files. I have dabbled with RAW and to date I have not been that impressed. I downloaded a freebie RAW converter that crashed all the time. I now have PS CS1 and with blown highlights on bird shots it failed to improve. The other pain was the Canon 300D it did not save RAW + jpeg. Because small birds move so quickly many shots are trashed and a quick method of proofing was required. So I reverted back to using jpeg. Now I have the 20D which will save RAW + jpeg I am willing to try RAW again but need to be convinced by example that it is really worth the effort and expense.

The jpeg process works by splitting the image into it’s chroma and luminance components ( colour & B/W). A transform is performed on these to determine how many levels of detail are in the image. This is similar to determining what sound frequencies exist in a piece of music. Depending on your jpeg encoder settings it first removes the high frequency components from the chroma and then as the level of compression is increased to extreme levels it starts to work on the luminance channel.
The remaining data is then compressed using standard lossless compression methods.

I have attached an example created in PS of low (fine)(top), medium and extreme (bottom) compression settings. I have zoomed the image to show the jpeg artifacts at each level. The low-fine setting usually uses co-efficients that give a useful compression before the jpeg artifacts start to show. I have compared the colour accuraracy between the original and the fine version and the difference in pixel values is about 0.5% with no noticeable overlap in colour.