Small size pictures utilize all sensor pixels?

[gialongo]

My camera (Canon Powershot S95) when set in the default 4:3 aspect ratio can produce pictures of the following sizes: 1) 3648x2736, 2) 2816x2112, 3) 1600x1200 and 4) 640x424
Are all sensors pixels used regardless of picture size or is only a subset of the sensor pixels used as needed to create a picture of the selected size?

[Dave Humphries]

I can't answer definitively, but effectively YES.

In other words, you won't get a small rectangle from the middle in 640 x 424.

Now whether it is averaging all the intermediate pixels (probably) or just using every 6th one to do this is largely irrelevant because; with memory not being the cost it was cameras were originally developed with this functionality, it serves no purpose in general photography to be skimping - always shoot at the maximum resolution; your "1)".

Welcome to the CiC forums from ...
Dave

PS
Could you do me a favour please?
Could you click Settings (right at the top),
then Edit Profile (on left)
and put your first name in the Real Name field
and where you are (roughly) in the Location field,
then click the Save Changes button below and to right,
this helps everyone give you more personal and relevant answers - thanks in advance.

PS2
I've got a Canon S100, great little camera I use every day

[george013]

My camera (Canon Powershot S95) when set in the default 4:3 aspect ratio can produce pictures of the following sizes: 1) 3648x2736, 2) 2816x2112, 3) 1600x1200 and 4) 640x424
Are all sensors pixels used regardless of picture size or is only a subset of the sensor pixels used as needed to create a picture of the selected size?

With the same framing?
In jpg?

George

[gialongo]

Yes, these sizes are for jpg.
Not sure what you mean by "With the same framing". If you mean the same aspect ration, then the answer is YES

[dje]

Are all sensors pixels used regardless of picture size or is only a subset of the sensor pixels used as needed to create a picture of the selected size?

Hi Gilbert

I can't be certain about what happens in camera but I think it is most likely that the camera starts with a full resolution image and uses a downsizing algorithm to produce a lower resolution image. It wouldn't save the full resolution image to the memory card, only the lower res image.

Dave

[george013]

Yes, these sizes are for jpg.
Not sure what you mean by "With the same framing". If you mean the same aspect ration, then the answer is YES

If the image/framing is the same, then the same sensor pixels are used.

Are all sensors pixels used regardless of picture size or is only a subset of the sensor pixels used as needed to create a picture of the selected size?

If the image/framing is different, a crop from the original, then that can be archived by just using a part of the sensor or by cropping afterwards. Is that what you want to know?

George

[dem]

In other words: when you change the resolution does the image magnification also change?

Does the 640x424 image have about the same angle of view as the 3648x2736 image or does it crop a small part of it? I bet it is the former.

[gialongo]

In my tests the picture content is the same regardless of image size as long as the aspect ratio is the same

[pnodrog]

In other words: when you change the resolution does the image magnification also change?

Does the 640x424 image have about the same angle of view as the 3648x2736 image or does it crop a small part of it? I bet it is the former.

Yes, it is the former. Actually, with CCD sensors all the pixels in a column need to be extracted as it is not possible to address and read random pixels. With most CIS (CIS = CMOS image sensors) it should be possible to read random pixels but I doubt that they do. I tend to agree with the assumption that the full resolution images are processed then software scaled to the selected resolution. Sensor designers will be concerned with getting maximum performance at full resolution and I doubt will take any consideration of performance or speed gains that may be possible when only lower resolutions are going to be stored.

Unless they adjust the PPI value the theoretical magnification will change but the sensor image area (and angle of view) will remain constant.

When memory was small and expensive reducing the image resolution to save memory was very occasionally useful. There is nothing more tragic than someone finding they have taken what would have been a prize-winning shot at an unprintable low resolution. The words uttered will be unprintable as well.

[gialongo]

I tend to agree with this

[Manfred M]

Yes, it is the former. Actually, with CCD sensors all the pixels in a column need to be extracted as it is not possible to address and read random pixels. With most CIS (CIS = CMOS image sensors) it should be possible to read random pixels but I doubt that they do. I tend to agree with the assumption that the full resolution images are processed then software scaled to the selected resolution. Sensor designers will be concerned with getting maximum performance at full resolution and I doubt will take any consideration of performance or speed gains that may be possible when only lower resolutions are going to be stored.

So far as I understand there is one exception to this and that is when cameras are used to record video. There is a significant amount of data that has to be processed, especially at high frame rates (50 / 60 + frames per second) and compressed using the appropriate video codec. Processing and throwing away most of the sensor data (a standard HD frame is ~ 2MB and a 4K UHD video frame is ~ 8MB) would require a faster processing engine, so I believe that there is something going on in the background to ensure that throughput is maintained..

[pnodrog]

So far as I understand there is one exception to this and that is when cameras are used to record video. There is a significant amount of data that has to be processed, especially at high frame rates (50 / 60 + frames per second) and compressed using the appropriate video codec. Processing and throwing away most of the sensor data (a standard HD frame is ~ 2MB and a 4K UHD video frame is ~ 8MB) would require a faster processing engine, so I believe that there is something going on in the background to ensure that throughput is maintained.

You a right Manfred. My remarks are only relevant to standard digital cameras at the current state of technology.

For specialised applications especially at lower pixel densities or with advanced CIS devices, I am sure there are quite a few options as to the most appropriate method of accessing each wells data and avoiding reading any that are not required. As we get more demanding about the performance of live view I think it will also have an impact.

I have seen some articles on digital/quantum image sensors that count photons captured or the number of auto resets that occur every time a small well is full. With this sort of an approach, they have a dynamic range only limited by counter size and noise. These sensors will be accessed almost like RAM and most will not require any A to D conversion. Sensor technology continues to develop at a staggering pace.

[george013]

Yes, it is the former. Actually, with CCD sensors all the pixels in a column need to be extracted as it is not possible to address and read random pixels. With most CIS (CIS = CMOS image sensors) it should be possible to read random pixels but I doubt that they do. I tend to agree with the assumption that the full resolution images are processed then software scaled to the selected resolution. Sensor designers will be concerned with getting maximum performance at full resolution and I doubt will take any consideration of performance or speed gains that may be possible when only lower resolutions are going to be stored.

Unless they adjust the PPI value the theoretical magnification will change but the sensor image area (and angle of view) will remain constant.

When memory was small and expensive reducing the image resolution to save memory was very occasionally useful. There is nothing more tragic than someone finding they have taken what would have been a prize-winning shot at an unprintable low resolution. The words uttered will be unprintable as well.

Magnification will always be the same. That solely depends on the focal length.

A FF camera that shoots as a DX camera does address a selected part of the sensor I think and can produce a RAW file of that part. Being addressed selective or cropped afterwards, the used sensor pixels for the image are different.
To change the image quality or resize it, first a raster image has to be produced from the RAW data.
Just my thoughts.

George

Magnification depends also on the distance of course.

[pnodrog]

We are using the term magnification in two different contexts. Yes, the image size on the sensor is dependent on focal length not resolution. However if the file is printed at 300 dpi without resampling/scaling the print size will depend on resolution.

[george013]

We are using the term magnification in two different contexts. Yes, the image size on the sensor is dependent on focal length not resolution. However if the file is printed at 300 dpi without resampling/scaling the print size will depend on resolution.

No, the file is not printed, just resized, in the question.
And all the sensor pixels are used for the image.

Magnification is used in relation to the sensor size. Enlarging is used to the print/screen size. Well, I think so.


George

[pnodrog]

No, the file is not printed, just resized, in the question.
And all the sensor pixels are used for the image.

Magnification is used in relation to the sensor size. Enlarging is used to the print/screen size. Well, I think so.


George

We have moved to the digital age now George.... The term enlarging is not really any longer applicable.... Within the context I wrote magnification most people will understand its usage.

[William W]

I am confident that Post @5 (Dave Ellis), in general terms, nailed the answer to the OP's question.

WW

[MrB]

...if the file is printed at 300 dpi without resampling/scaling the print size will depend on resolution.

ppi

[george013]

Before starting another topic it's not an issue for the question. That's resizing. And that's done on the full rgb raster image.

George

[pnodrog]

ppi

The file would be in PPI (if defined) the printing DPI.

Personally I am far more interested in concepts rather than arguing about semantics.