Auto-focus sensors: high precision vs. regular testing

Started Apr 4, 2010 | Discussions thread
hjulenissen
Senior MemberPosts: 1,617
Like?
Re: Using JPEG file size as an indicator of sharpness
In reply to mmullen, Sep 27, 2010

mmullen wrote:

hjulenissen wrote:

mmullen wrote:

JPEG file size as a measure of image sharpness is well established as being an accurate gauge.

Well established as in there are peer-reviewed papers doing this? Or knowledgeable people that you can refer to?

I'm sure there are peer reviewed papers that acknowledge that JPEG file size varies with image detail but I don't know of any off-hand. There are also a number of knowledgeable people that recognize this fact and hold it to be true based upon repeated observations.

Of course JPEG file size can vary with image detail. Using JPEG file size as a direct indicator of sharpness is taking it to another level, would you not say? Can you mention some of these knowledgeable people, perhaps guide me to forum posts or websites?

"If the compression ratio can somehow be guaranteed (irrespective of image content), then the resulting file size can also be guaranteed. There are some situations where optimizing for a particular file size may be an advantage.

So far, I have come across the following cameras that exhibit this feature. Note that others may exist, so if you spot one, please let me know.

That's an interesting article but it is of no relevence to this discussion because it concerns compression schemes that use variable compression depending upon image detail. The images I used in my test were compressed into JPEG using the 7D on the same quality setting and thus the same quantization table for each image.

I am sorry but I think that it is relevant. Did you manually inspect the quantizatin tables? Those cameras mentioned seems to use varying quantization without user intervention. It is hard to exclude that cameras used in your test did strange stuff to their jpeg files that makes sense for human observers, but does not make sense for your test. Even if the quantization table stays constant, there are possibilities to trade size vs quality (deleting small coeffs).

Reply   Reply with quote   Complain
Post (hide subjects)Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow