Strange images from 20D, have you noticed?

[tannoy]

Hello,

I have been very happy with the overall performance of the 20D but something Pekka S. said a few months ago has bubbled up top the surface in a few of my images.

I shoot raw with parameter 2 90% of the time.

What is happening is that very faint detail is being lost by what appears to be the noise reduction algorithim in the camera. A photo I recently took of the Golden Gate exhibits this clearly(shot at iso 200). One of the small cables that runs above the main support cable is visible everywhere except where there was nothing but blue sky. In this area it has been severly compromised (almost erased) and looks remarkably like what happens if I use too much noise reduction in Noise Ninja.

Image can be viewed here:

http://homepage.mac.com/kavanaghdarrinsean/PhotoAlbum12.html

Has anyone else had this problem or noticed the phenomenon. Next question; is does the 1D Mk2 have the same problem?

Cheers,

Darrin

[PhotosGuy]

1st, You have to have the image online.
Then, click on the Img button, enter the image location & click on the Img button again so it looks like this.

image location

[booggerg]

Why does the image look so bad to begin with? the trees in foreground is all fuzzy and it just looks bad over all. Is that JPEG heavily compressed?

[tannoy]

Tree is slightly out of focus, but no it is not heavily compressed, the TiFF exhibits the same behavior on the cable.

Cheers,

Darrin

[ScottE]

Notice that where the cable is about the same shade as the sky and lighter than the dark red. It is possible that the difficulty comes because there is less contrast, not because of noise reduction. It is much easier to resolve a light object against a dark background than a light object against a light background.

Since there is low contrast the camera is going to struggle with resolving the cable.

Any error in focus, movement of the camera or softenss of the lens is going to add to the problem. The edges of other objects in the photo against the sky are not exactly crystal sharp in the rest of the image. Any softness in an object as narrow as the cable are also going to make it more difficult to resolve.

Those are just my guesses.

[tannoy]

Thanks for the input, did you notice the strange look of the pixels around the cable?

Darrin

[shiningstardv]

tannoy,

Just to let you know -- .Mac is not a good place to upload photos for comparison and examination, simply because iPhoto & .Mac (or just .Mac, depending on how you uploaded) compress your images so much. It's difficult to see fine detail when the JPEGs are highly compressed. I can;t quite seem to see the problem your talking about...is it in the main cable or one of the smaller support cables?

[ScottE]

Thanks for the input, did you notice the strange look of the pixels around the cable?

Darrin

If it is a resolution problem, the blue near the cable is going to be effected the same way the cable is effected.

It is interesting to note that the small cable above the other support cable shows up black against the blue sky because it is shaded. There is no problem resolving that because there is more contrast against the light blue sky.

[davidwegs]

What ScottE said.

BTW, if you shoot RAW the parameters mean nothing in camera. They are only applied to the JPEG's produced by the camera. Except for the B&W setting and even that is still recorded in color when shot RAW.

Hope that is also of some use.

[drisley]

Like David said.
Raw is unaffected by the Parameters setting.
BTW, what software are you using to convert the Raw. If it is DPP 1.1, that is more than likely the problem you are seeing.
There is a problem with the demosaic algorithm used in DPP to cause the exact problem you are showing.
I've seen the problem, and so has THIS REVIEW (http://www.outbackphoto.com/artofraw/raw_14/essay.html)of DPP.
The other Canon software included with the 20D, EVU, actually uses a completely different engine. This problem disappears with EVU.
Too bad. EVU produces much nicer images than DPP, but DPP is SO much nicer to use.

[tannoy]

Thanks for all of your input, I still have my suspicions about the noise reduction and will need to do a bit more experimentation. The reason I suspect this is the shape of the noise on this image and some others I have, it is inconsistent based on the amount of detail in the image (much like Noise Ninja) and appears to sqaunder detail but only under specific circumstances.

Drisley this was developed using the EVU not DPP but thanks for the heads up on DPP, what a shame.

Cheers,

Darrin