Nikon D3200 A cheap NEX-7?

Started Jun 17, 2012 | Discussions thread
Aku Ankka
Aku Ankka Contributing Member • Posts: 591
Re: ah, to you pixel binning is ONLY in "hardware"

RussellInCincinnati wrote:

However my repeated interest in a pixel-binning hi-res camera, is to have the option of capturing low-res, nice-color (though perhaps not as nice as raw files, but so what? That's not the point) JPEGs at a higher frame than the storage media could tolerate with raw files .

The key here is that you do not understand how the low-resolution JPGs are done in the camera today - they have better image quality than your proposition would give. They are not binned in anyway, but downsampled from the full size images. This gives noticeable detail advantage over your method.

Aku: No, many users find them acceptable and convenient, but they have significant advantage over raw: they are much smaller. TIFs are not and may even be much bigger.

You can't imagine someone wanting simple TIFFs out of the camera, rather than hardware-dependent raw files, even if they're huge.

Did I say so? Where? Stop lying about what I say!

explaining some of your frequent name-calling.

Calling you a liar is not name-calling, but stating a fact.

I've expressed interest in pixel binning to improve frame rates

Indeed - and I replied on that. When I did that you decided that sensor, which is the limiting factor, is not part of the discussion, but instead something else. Your agenda it not interest in facts, but winning an argumet at any cost.

, never said that pixel-binning can do something that couldn't be done in more leisurely post-processing.

Not post processing, but processing. You keep insisting how your method will some mystical way create better image quality and conventional downsapling of the data, you know like it is done for the low-res JPGs.

Aku: There are not shortcuts. You can create far superior output image regardless of the conditions from all the data, than some of the data.

Software pixel binning before camera storage is in contrast looking at all the data . But then extracting some of the resolution (and a lot of the color info) out of all those un-needed-for-every-purpose raw pixels, and then quickly writing out a low-resolution summary of the input data to achieve higher frame rates.

You are just describing what is downsampling , no matter how you call it. The current cameras create superior image quality low-res images compared to your proposition.

Speed wouldn't be improved significantly by your method either as the limit for high frame rates is in transferring the data out of the sensor (including AD conversion). Your method would speed this up by a small factor (only for the greens), while it would require significantly more complex sensor which in turn would have lower performance compared to sensor without the extra circuitry.

Aku: There is one advantage to binning and that is to bring down the data transfer rate

Hmm, which you did not acknowledge previously.

You keep on lying and not understanding what I write. Your software trick does not decrease the data transfer rate in any way, shape or form. The relevant part is moving data from the sensor in digitized format to the image processor.

When writing to the buffer your method offers no size advantages over conventional downsampling algorithms either.

Aku: this is why it's used for video and live view...(though in this binning lots of pixels are just ignored due to the problems of creating a good binning for Bayer filtered sensor - this is why the noise levels are high for live view).

Ah,these thoughts make sense to me now, once I understand your definition of binning as hardware-only, agreed.

Excellent! Btw, you are the first person I have even heard talking about binning as software solution - in the relevant literature binning is a hardware operation. The software equivalent is called downsampling.

Russell: I see, reducing resolution by 75% will result in a loss of resolution. Forewarned.

Aku: No, you don't see it. By downsampling to 2*2 superpixels, you lose 50% of resolution.

So getting rid of 75% of the input pixels you only lose 50% of the resolution. We're obviously talking about different measures of resolution. I see, you are talking about linear lines-per-inch-in-one-direction resolution, OK.

Resolution is a linear metric. Line pairs per unit of length/width.

Aku: Do you not understand that you propose that in some mystical way you gain in image quality by throwing information away?

Do you not understand there could be storage and framerate advantages to throwing away some unnecessary resolution info in the field, while keeping some of the lower-noise info--without constantly referring to those advantages as "mystical"?

The reason why I keep on repeating this is that what you propose is in no way an improvement in the current state of affairs. Your method is inferior in quality (and likely in size as well, especially for noisy images) compared to the downsizing methods used in contemporary cameras.

You can test this yourself if you want to:

  1. Pick a suitable raw file

  2. decode it with dcraw document mode (no demosaicing)

  3. Do your simplified demosaicing on this file

  4. Compare the result to a normal downsampled image (ie. raw-> tiff-> downsample-> slight sharpening)

Aku: The raw data does not improve in some mystical way if you do the simplest demosaicing there is.

Inflammatory and uninformative.

No - instead you don't excel at reading comprehension. You've been very vague on what you're talking about and shifting focus constantly. In order to respond I have to write in very many topics.

However, am I right in understanding that in this discussion you are only interested in small JPGs out of camera with maximum quality and minimum size and maxmum frame rate?

-- hide signature --


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