I try to understand what happens now in digital imaging and what changes will be most beneficial for a new generation of cameras but I have to admit there are too many factors involved and we just have to rely upon the "magic" Nikon works with all the different variables available to it.

Hi all,

TTJ for the win...

There aren't easy answers, unless you'll settle for "magic".

In my Digital Imaging class, the students would try to guess which images was 50KB and which was 100MB. Then we tossed a coin for which was which.

A coin toss was usually more accurate.

It was a good lesson on compression.

My best,

Mike

There is really not that much manipulation of pixel data as one may think especially with raw. The real next IQ jump will be when there is another color bit depth increase with a bump to 16bit color. Currently that is one of the few items MF cameras have over FX cameras.
One hell of a processor will be needed for that - bit depth info grow exponentially instead of the opposite of MP and Image resolution. If this will be it, who knows. One thing for sure is that there are many, many components that go into increasing IQ, much more than any of us know.

So the Large, Medium and Small sizes are all software manipulation of pixel data and not any "binning." That is what I thought was happening.

I read Nikon was producing a new EXPEED processor which will be able to handle more data faster. Perhaps such a processor will be able to offer additional automatic software improvements to IQ. I have seen some comments that it takes double the pixels to really see a significant difference and the new D4 is likely to have only 50% more pixels. But I have also seen comments that the new D4 will offer significant improvements in IQ. So where does that significant additional IQ come from? Not lenses, not 50% more pixels. Perhaps the largest contributor to increased IQ will come from a new and more powerful EXPEED processor running new software.

I am not stuck on "binning." It was just an idea for better high ISO IQ. Now I am more inclined to think increased processing power will yield the greatest improvement in IQ.

Hopefully, we will soon have some announcements from Nikon touting a new generation of camera bodies with improvements in IQ due to X,Y and Z. I am betting x will be a better sensor and Y will be a faster processor and Z will be new software manipulation of the pixel data.

It is all compression, color gamut and bit size. I just compresses and re-sizes the images like PS would. It does not "turn off" pixels on the sensor.

donaldejose - You really are stuck on "binning" being the end all solution aren't you?

As far as I understand the question it is about image size, not file size. The D700 has got three image sizes called S, M and L where S (2128 x 1416) is exactly half the size of L (4256 x 2832). So in that case Nikon could only use every second pixel on the sensor or bin them if it was a grey value sensor.

But it is a color sensor, so binning does not make sense. I am not sure but I think that Nikon uses some kind of Bayer pattern sensor. In that case if you would use binning you would throw the information from two green, one blue and one red pixel together. If you would only use every second pixel you would also loose information (e.g. in one column you would only get the green pixels), so again you would not gain anything.

The M size is in between, so the image has to be down-sampled using the on-board processor. And I am pretty sure that they do the same thing for the S size as well.

Would a "better" processor improve image quality? I think so. I remember another thread (can't find it right now) where someone said that running some older RAW images through current software (which I think is similar to a "better" processor) showed better results than some years ago. A faster processor with newer software could improve IQ by using more sophisticated algorithms for color reconstruction from Bayer, NR for long exposure etc.

Just my thoughts on this...

Digital cameras can save their images in two different ways - either 'raw' or compressed.

Raw images record the exact colour of every pixel in the original image. Some savings can be made by noticing that adjacent pixels are the same colour, but otherwise each and every pixel must be recorded and the file sizes are therefore pretty large.

Compressed images take advantage of similarities between adjacent pixels to store a rough estimate rather than an exact colour.

There can be some fairly mind-bending maths involved, but the result is that you can get a fairly accurate representation of an image without storing each and every pixel precisely.

This means that you get smaller file sizes. The exact size will depend on how much detail the original picture contains, and how accurate the stored version is required to be.

The same image can be stored with higher compression (lower accuracy) and take up much less space on disk. The penalty is that higher compression introduces 'noise', where the clever maths tries to guess at the exact colours of each pixel and gets it a little bit wrong. Noise typically appears as speckles, particularly around sharp edges in the image and areas of high contrast.

hope this helps.

Really, only Nikon knows what they do, and it's unlikely that they will share the information with anyone, especially not end-users.

The best way to find out is to do your own tests. Shoot a test chart or a test scene with the various size settings, and compare the smaller images with images you resize on your own. If they look close enough, than that should be fine. I'd also repeat this test at high ISO, to have a look at what impact the size has on noise.

No, sorry. I know what "binning" is. That last thread was about how "binning" could be a way to gain larger pixels on a DX sensor. This thread is really not a "binning" question as I think no cameras currently use "binning" and to my knowledge none are expected to soon use "binning."

What I don't know is how the camera takes an image in the different sizes? That was my question for this thread.

Is a reduction of maximum pixel use done at the senor level by just using less pixels to capture the image (either by simply not using some or by combining some (like binning - which I understand is not done) or is the reduction of pixels really a software event: i.e. The camera uses the full pixels on the sensor to capture the image and then uses software to compress that image before being stored on the SD card. My guess is that the senor always "sees" the large size image and generates an output from each pixel light strikes on the sensor but software reduces the pixel count to medium or small size for storage. Is that what is happening? That was my question.

Now for an implication of my question.
If it is a software manipulation of data from each pixel on the sensor then "binning" could be done in the software if you have enough processing power. I am starting to think many different things can be done as software processes the data from each pixel and these software manipulations of the data may be more important than the amount of data streaming off the sensor. Hence, a new faster dual core EXPEED processor may make a more significant contribution to improved IQ than more megapixels in the sensor.

Many of us tend to think improvements in IQ come mainly from more pixels on the sensor. Perhaps a new EXPEED processor will be more important than more pixels.

Didn't you already get an answer to this in your last "binning" thread?
http://nikonrumors.com/forum/topic.php?id=3793

We all know the menu allows you to chose image size.
We all know the different sizes contain different amounts of megapixels.

But, how does it work? Is the camera using all the same pixels in the sensor regardless of how you set the size and then just compressing the image into fewer pixels to get a smaller size file? Or is the camera somehow using fewer pixels on the sensor in the first place as it captures the image?

Here is why I wonder what is going on. If the camera is using, for example, one fourth of the sensor pixels to capture a small size image why not "bin" four pixels together (combine their data into one) and thereby achieve the benefits of greater low light sensitivity which can be obtained from larger size pixels because now you have four times more light falling on one "binned" pixel?

If this were true a person could decrease noise at high ISO by switching to the small image size. Just an idea and wondered if it could work.