This article should be read by everybody

Started Dec 12, 2013 | Discussions thread
TrojMacReady Veteran Member • Posts: 8,729
Re: Both sides have a point.

artlmntl wrote:

The article posits that DxO testing gives an accurate measure of noise in images. DxO does this by downsampling to an arbitrary size, 8MP. The article indicates that in a downsampled image, sensors of the same surface area will give similar results.

I think this is true - I've done some tests with sample RAW files from different cameras and when I downsampled them to the same arbitrary size (for my purposes, 3000x4500px while evaluating FF sensors), the resulting images were very much the same, even at the pixel-peeping 100% view - there was nothing to pick between them. I saw variations in white balance and color rendering, nothing that would be a problem in post processing.

Anybody can download and resample a bunch of sample files, so it's not really important to provide samples. Besides, people know this already.

The thing is, if we buy a large-resolution camera, that indicates we plan to crop into the pictures or some way use the pictures larger than we would the lower resolution image. If not, why bother? And that is the point where I think the article comes up a little short.

Two images shot with cameras with same size sensor (1", APS-FF, etc) downsampled to the same size will look similar as long as all the other shooting parameters are about the same. But when you start using the larger resolution images to actually provide more resolution, you also see the noise more clearly. This is your point, I think.

And this is a point that is subject to an arbitrary factor: what we subjectively find acceptable for a given output size. As you move up the ISO scale, no doubt the benefit of printing larger or having more detail/smoother files at a given output size, will diminish, just like it did going from 4 to say 10 MP. And diminishing benefits are not the same as downsides (buffer, shot to shot and storage issues would qualify for that). But as for example shown with the Nokia Lumia cellphone cameras, high MP Bayer sensors can have clear benefits at smaller output sizes too, when using a good resampling algorithm. Less sharpening artifacts, more NR options, cleaner transitions and more color information (Foveon like but without the issues in the bottom red layer/channel).

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