M11 and Pixel Binning

Started 11 months ago | Discussions thread
OP joslocum Senior Member • Posts: 1,882
Re: M11 and Pixel Binning

Tom Schum wrote:

joslocum wrote:

Tom Schum wrote:

joslocum wrote:

The M11 is able to record RAW (DNG) images in 60MP, 36MP, or 18MP file sizes.

My limited understanding of Pixel Binning is that the sensor somehow combines four pixels into one "super" pixel and this reduces the file size while attempting to maintain image quality.

I've been confused about M11 image sizes also.

If we "pixel bin" the 60mp sensor into Bayer quads (a square arrangement of one blue, one red, and two green sensels) the output image would be 1/4 the original 60mp = 15mp. Noise and dynamic range specs are not necessarily improved by this method, but microcontrast might be improved.

15mp is not one of the image size choices. So, I don't think it should be called "pixel binning".

Perhaps the camera is internally resizing the images to smaller pixel counts. If so, noise and dynamic range specs would be enhanced in some proportion related to the amount of downsizing. Lots of cameras do this with in-camera JPGs, and few do it also with raws. Looks like Leica is doing it with the raws. This is a good thing.

Crop modes are a different story.

That's exactly what I was thinking Tom.

60 > 30 > 15

Obviously I'm missing something.

Here is the video I watched about the Leica M11 Dynamic Range and Pixel Binning:


Note: FF to about 13 minutes into the video for pixel binning. Also note that the round EVF mounted on top of the M11 is for an M10. He explains why in the video.

Really excellent video. The 60mp sensor seems up to the task with the Summicron 35mm lens too.

I have a Sigma fp L, and I believe the basic sensor is the same as the M11 except for the microlenses and the color filters.

To get a 30mp image I have to reduce the pixel dimensions by the square root of two. That is, a width of 9528 pixels becomes 6737 pixels.

To get a 15mp image I have to reduce the pixel dimensions by two. That is, a width of 9528 pixels becomes 4764 pixels. In this case I can see how it would be possible to "bin" two pixels into one in the x dimension and "bin" two pixels into one in the y dimension, so that each pixel in the 15mp image includes the effect of four pixels in the 60mp image. Noise is cut in half in the 15mp image because of the averaging of four pixels.

I've always thought that binning has to be with integral numbers of pixels. Maybe Leica is thinking differently.

For example with the 30mp image we are "binning" 1.414 pixels in x and y, to get an image with half the original pixels. Noise will be one over the root of two = 0.7071x the noise of the original image.

Now we can use "bin" to apply to anything at all. Seems counterproductive to me, but...

The full size 9528x6328 (60.29mp) image gets downsized to 7413 x 4928 (36.53mp) or 5272 x 3498 (18.44mp).

Theoretically the noise of the downsized image is the square root of the ratio of pixels. For the medium M11 image the noise is the square root of 36.53/60.29 = 0.778x relative to the large image. For the small M11 image it will be the square root of 18.44/60.29 = 0.553x.

And for the truly binned 15mp image (not one of the M11 options) it will be 0.5x. In terms of power (1/2 the power), this is 3dB down. If dynamic range is 14 bits we are talking about one bit less, 13 bits. I'm a little rusty with this stuff so I might be telling you wrong here.

I'm guessing it would be difficult to clearly see a noise reduction in the M11 small image, relative to the M11 full size image, so the reviewer in your linked video failed to see it.

I really enjoy the videos from "mathphotographer".  I know absolutely nothing about the guy other than his first language seems to be German, as would be the case if he is from Zurich.  It remains unclear to me if he shoots video in his basement or if this is a shop somewhere.  But I like the clean look of his surroundings and his videos make sense and are well done in most cases.

Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow