Mel Gross
Offline
• • •
Upload & Sell: Off
|
Lotusm50 wrote:
Within the samples referred to in the above link, is comparable full-resolution RAW images from the Sony a900 and Nikon D3 taken at 1600 ISO. I downloaded the RAW files and did a straight auto conversion to both using Adobe Camera Raw 4.6 (latest version accommodating the a900). Even downsizing the a900 to an equivalent resolution as the D3 images, the a900 showed significantly more noise, and less detail. Now I understand that the Nikon has had some measure of noise reduction applied in-camera to the RAW file, and the a900 probably did not. Adding noise reduction to the a900 might have helped the noise, but it could only have made the level of detail available in the a900 file even worse. I'll redo the conversion on my desktop to duplicate and confirm, and I'll see if I can upload illustrating images (currently I just have my laptop).
...Show more →
The Nikon sensors themselves are supposed to have much less noise before noise reduction.
It's known that the Sony has heavy noise reduction applied before A/D conversion, and again after A/D conversion, and the pre A/D noise reduction can't be turned off. Unless Sony can do this with a software update to the firmware, it will always have heavy noise reduction in the RAW files.
So, your results simply show that the Sony has horrendous noise at the sensor, and possibly poor control in the processor.
From the Luminous Landscape field review which I posted before (compared to the Canon 1Ds mkIII):
The Sony specs also indicate a two step noise reduction system, one before a/d conversion and one after. This first step is not user controllable and I see very subtle smudging of fine detail which may be connected to this.
These images show this concern in the grass which looks slightly smudgy on the Sony and also in the cobblestones at the waters edge where the Canon clearly looks more textured...
Comparing the two camera with all NR and Dynamic Range Optimising turned off and at base ISO, the noise levels are very good, in fact very similar to the Canon images. This is pleasing, although some fine detail, as mentioned above, may be masked by the in-camera NR......Show more →
While he seems to have found numerous problems with the files as compared to the Canon, he still thinks it's fine, considering the price difference.
When compared to the Nikon noise, both the Sony and Canon are definitely wanting.
The 5D mkII should be much better in that regard.
Link:
http://www.luminous-landscape.com/reviews/cameras/a900-nr.shtml
I've never seen convincing evidence that downsizing an image file reduces the noise. It's different when the camera itself does this, because it then uses the sensing sites themselves to reduce noise. We can't do that. The noise is already there. Reducing the image file size simply spreads the noise around, often making it look worse.
In fact, given equal noise at the pixel level, the file that's larger (from more pixels in the sensor) will appear to have less noise, given equal print sizes. If the larger file has somewhat more noise, it might look equal, again given equal picture sizes.
That's simply because the pixel, and therefor the noise, is smaller, making it less obvious at normal viewing distances. It tends to blend better.
It's one reason why really high resolution medium format has high IQ. I can see the noise from the Leaf backs I test, but on print, it's hard to see without a glass.
|
