Noise 100% 40D 7 G9 - 100 & 800 ISO

We all know P&S's aren't gonna compare with a decent DSLR, but just how much difference is there? The following shots take with G9 and 40D/17-40mm 4.0 L combo. All shots F4.0. on tripod. All are 100% crops ... G9 lookin' pretty crummy at 800 iso in comparison (yes, I know price differences, etc!) but you can also see real world usage on web shots below, and things look better.

So let's see the whole original shot reduced down to 800px and see what the difference is in real world use on the web ... again, only PP is white balance in RAW. However, I've used 200 iso for one set as that would be my own maximum for a P&S in regular use and I was curious. All look usable tho' clear quality difference in favour of the 40D - this would be even more marked if the shot was sharpened of course.

In ACR, try these settings on the Details page: 60, .8, 30, 0, 80, 100 (top to bottom) as described here:
http://forums.dpreview​.com …orum=1010&threa​d=25612433

Well that G9 certainly looks better at ISO800 than my Sony H5 superzoom did. I never took that thing past ISO400 if I absolutely had to.

Hi condyk,

Excellent post - thank you for taking time out to do it. The marriage of a G9 to partner my 40D looms closer.

Bob.

GordonSBuck wrote in post #4468895
In ACR, try these settings on the Details page: 60, .8, 30, 0, 80, 100 (top to bottom) as described here:
http://forums.dpreview​.com …orum=1010&threa​d=25612433

I don't use ACR so if you want to try the 800 iso shot through those settings and then post results then PM me your email address and I'll send it to you. These shots only had WB from RAW.

Hey Gordon... here's an interesting note for you...

I chose an ISO 800 image from the G9, and then applied those posted noise settings to the CR2 RAW file in ACR 4.3.1. I also seperately ran the "adaptive noise" filter (setting 10) in Canon's RAW converter that came bundled with Zoombroser & the G9. In both cases I outputted a 16-Bit TIFF and viewed both shots in PS CS3.

I gotta tell you - the Canon adaptive noise filter seemed to do a better job. Now maybe its because Canon hasn't release the SDK for the G9, but it seems to me at this stage that the "in-house" noise reduction created by Canon for the Canon G9 might do a better job than Adobe's ACR.

I find it hard to believe to, and I am going to repeat my tests tonight on additional images. I will post what I find in a new topic in this forum.

BTW - I also ran the image straight through the RAW converters with all standard settings, and them applied noise reduction using the latest Noise Nija plugin. It too seemed to pale.

openspace wrote in post #4472437
Hey Gordon... here's an interesting note for you...

I chose an ISO 800 image from the G9, and then applied those posted noise settings to the CR2 RAW file in ACR 4.3.1. I also seperately ran the "adaptive noise" filter (setting 10) in Canon's RAW converter that came bundled with Zoombroser & the G9. In both cases I outputted a 16-Bit TIFF and viewed both shots in PS CS3.

I gotta tell you - the Canon adaptive noise filter seemed to do a better job. Now maybe its because Canon hasn't release the SDK for the G9, but it seems to me at this stage that the "in-house" noise reduction created by Canon for the Canon G9 might do a better job than Adobe's ACR.

I find it hard to believe to, and I am going to repeat my tests tonight on additional images. I will post what I find in a new topic in this forum.

BTW - I also ran the image straight through the RAW converters with all standard settings, and them applied noise reduction using the latest Noise Nija plugin. It too seemed to pale.

Wow, that's amazing! I'm very interested in noise reduction for the G9, especially at ISO 800, so please double check and let us all know your findings. I too have tested several images, software and techniques. Frankly, the Canon software has not come out very well in my comparisons.

I'm almost certain that Adobe does not rely on the Canon SDK for their raw processing. Those ACR settings, developed and reported by Thomas Niemann (see the link), have worked out best for several different images.

Is it possible that the sensor in the G9 generates noise not in a predictable mathematical pattern, but a signature pattern specific to the engineering of the G9 sensor? If this is the case, then those who know that signature pattern would be in best position to address this noise in their software. This also is the information I would guess would be in a Canon SDK - i.e. signature algorithms for addressing signature noise.

If Adobe doesn't use Canon's SDKs then perhaps ACR is not optomized to best handle signature noise patterns for individual sensors.

openspace wrote in post #4473304
Is it possible that the sensor in the G9 generates noise not in a predictable mathematical pattern, but a signature pattern specific to the engineering of the G9 sensor? If this is the case, then those who know that signature pattern would be in best position to address this noise in their software. This also is the information I would guess would be in a Canon SDK - i.e. signature algorithms for addressing signature noise.

If Adobe doesn't use Canon's SDKs then perhaps ACR is not optomized to best handle signature noise patterns for individual sensors.

Canon does not manufacture the G9 sensor but they certainly should know a lot about it when combined with the G9. I don't know if the Canon software does noise reduction specific to the different sensors that they use. It would make sense and be great news for all of us if the Canon software provided the best noise reduction for the G9 but deep down inside I'm pretty sure that better noise reduction is available elsewhere. The thing about noise reduction is that the final image quality is a compromise between noise and resolution. I'll be doing some additional testing while watching for your results.

great post! many of user ask the question like this. This is the good answer