Image Size

**tcwilson** · 11-04-2010, 08:08 PM

When I change the image size in my K-x from 12 megapixels to 10, to 6, etc., what is the camera actually doing? Does it record fewer pixels from the sensor? If so, which pixels does it use? Does the different pixel pattern affect the image (aside from the obvious loss of resolution) Or, is the change in image size basically a software function where the camera takes several pixels and processes them as one?
Thanks.

Capslock118 · 11-04-2010, 08:17 PM

I don't know.

But If I had to guess, I would not say it's happening at sensor.

I would guess that the picture is taken, then software resizes the photo, as you would in photoshop or aperture or gimp.

hcc · 11-04-2010, 10:52 PM

My understanding is that the camera record at full resolution (eg 12 Mp) and then process/convert the RAW data into JPEG. I do not know the algorithms, but there must be some computations involved from 12 Mp to 10 Mp.

This is why the K-5 has a function to convert a JPEG into RAW as long as the original data are in the camera buffer.

I cannot guarantee that I am right however for K-x and K-7, but I am probably not far.

**tcwilson** · 11-05-2010, 05:53 PM

Thanks for the information. If the camera really combines pixels via software to get the resized image that raises an interesting question about noise, since the pixels are analog sensors.
I know an electrical engineer could straighten me out on this, but here's my theory. For the sake of example, we set the camera to produce a 6 mp image instead of 12mp By combining two real pixels into the new single "virtual" pixel" you would double both the signal and the noise. So the S/N ratio would remain the same. But since you have now doubled the total signal, you can attenuate it lower, also lowering the noise below the threshold of detection.
Now in experimenting with the different settings, this clearly isn't the case. But if engineers could figure it out, it would be neat trick. Just imagine the low light response you'd have at a 2mp setting.

egordon99 · 11-09-2010, 10:20 AM

Originally posted by tcwilson

Thanks for the information. If the camera really combines pixels via software to get the resized image that raises an interesting question about noise, since the pixels are analog sensors.
I know an electrical engineer could straighten me out on this, but here's my theory. For the sake of example, we set the camera to produce a 6 mp image instead of 12mp By combining two real pixels into the new single "virtual" pixel" you would double both the signal and the noise. So the S/N ratio would remain the same. But since you have now doubled the total signal, you can attenuate it lower, also lowering the noise below the threshold of detection.
Now in experimenting with the different settings, this clearly isn't the case. But if engineers could figure it out, it would be neat trick. Just imagine the low light response you'd have at a 2mp setting.

And that's pretty much what you get when you downsize your image (in Photoshop or whatever) to 1000-pixels wide, the noise doesn't appear to be "as noisy"

The camera most certainly uses a similar process (software algorithm) to downsize the image prior to saving it on the card.

ezza · 11-09-2010, 06:26 PM

Originally posted by tcwilson

Thanks for the information. If the camera really combines pixels via software to get the resized image that raises an interesting question about noise, since the pixels are analog sensors.
I know an electrical engineer could straighten me out on this, but here's my theory. For the sake of example, we set the camera to produce a 6 mp image instead of 12mp By combining two real pixels into the new single "virtual" pixel" you would double both the signal and the noise. So the S/N ratio would remain the same. But since you have now doubled the total signal, you can attenuate it lower, also lowering the noise below the threshold of detection.
Now in experimenting with the different settings, this clearly isn't the case. But if engineers could figure it out, it would be neat trick. Just imagine the low light response you'd have at a 2mp setting.

Fujifilm did something along those lines in a few of their P&S cameras, called EXR. dpreview talk a bit about how it works in this review: Fujifilm FinePix F200EXR Review: 2. EXR technology: Digital Photography Review

I owned the F200 for a while and I can confirm that the system does work well, although I found it more useful for giving me extra dynamic range than reducing noise.

**tcwilson** · 11-10-2010, 07:29 PM

Thank you for the information. I appreciate your taking the time to reply.

11-04-2010, 08:08 PM	#1
tcwilson Site Supporter Join Date: Nov 2010 Location: Cincinnati, Ohio Photos: Gallery Posts: 70	Image Size When I change the image size in my K-x from 12 megapixels to 10, to 6, etc., what is the camera actually doing? Does it record fewer pixels from the sensor? If so, which pixels does it use? Does the different pixel pattern affect the image (aside from the obvious loss of resolution) Or, is the change in image size basically a software function where the camera takes several pixels and processes them as one? Thanks.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Image Size vs Document Size vs Resolution vs Resampling vs ... AHHHH!	veezchick	Digital Processing, Software, and Printing	13	08-02-2010 03:57 PM
Image size	Jacos	Troubleshooting and Beginner Help	15	04-11-2010 02:47 PM
JPEG Image Size	hcc	Troubleshooting and Beginner Help	21	04-10-2010 08:44 PM
1.53× vs. 1.54× (or even 1.57×) and actual image size	mattdm	Pentax DSLR Discussion	0	05-23-2009 06:44 PM
Photo image size from my K200D	fractal	Troubleshooting and Beginner Help	7	01-11-2009 09:51 AM

11-04-2010, 08:17 PM	#2
Capslock118 Senior Member Join Date: Apr 2010 Location: New Haven, CT Photos: Gallery Posts: 173	I don't know. But If I had to guess, I would not say it's happening at sensor. I would guess that the picture is taken, then software resizes the photo, as you would in photoshop or aperture or gimp.

11-04-2010, 10:52 PM	#3
hcc Pentaxian Join Date: Nov 2009 Location: Brisbane, Australia Photos: Gallery \| Albums Posts: 4,005	My understanding is that the camera record at full resolution (eg 12 Mp) and then process/convert the RAW data into JPEG. I do not know the algorithms, but there must be some computations involved from 12 Mp to 10 Mp. This is why the K-5 has a function to convert a JPEG into RAW as long as the original data are in the camera buffer. I cannot guarantee that I am right however for K-x and K-7, but I am probably not far.

11-05-2010, 05:53 PM	#4
tcwilson Site Supporter Join Date: Nov 2010 Location: Cincinnati, Ohio Photos: Gallery Posts: 70 Original Poster	Thanks for the information. If the camera really combines pixels via software to get the resized image that raises an interesting question about noise, since the pixels are analog sensors. I know an electrical engineer could straighten me out on this, but here's my theory. For the sake of example, we set the camera to produce a 6 mp image instead of 12mp By combining two real pixels into the new single "virtual" pixel" you would double both the signal and the noise. So the S/N ratio would remain the same. But since you have now doubled the total signal, you can attenuate it lower, also lowering the noise below the threshold of detection. Now in experimenting with the different settings, this clearly isn't the case. But if engineers could figure it out, it would be neat trick. Just imagine the low light response you'd have at a 2mp setting.

11-09-2010, 10:20 AM	#5
egordon99 Inactive Account Join Date: Dec 2006 Location: Outside of Philly Posts: 1,561	Originally posted by tcwilson Thanks for the information. If the camera really combines pixels via software to get the resized image that raises an interesting question about noise, since the pixels are analog sensors. I know an electrical engineer could straighten me out on this, but here's my theory. For the sake of example, we set the camera to produce a 6 mp image instead of 12mp By combining two real pixels into the new single "virtual" pixel" you would double both the signal and the noise. So the S/N ratio would remain the same. But since you have now doubled the total signal, you can attenuate it lower, also lowering the noise below the threshold of detection. Now in experimenting with the different settings, this clearly isn't the case. But if engineers could figure it out, it would be neat trick. Just imagine the low light response you'd have at a 2mp setting. And that's pretty much what you get when you downsize your image (in Photoshop or whatever) to 1000-pixels wide, the noise doesn't appear to be "as noisy" The camera most certainly uses a similar process (software algorithm) to downsize the image prior to saving it on the card.

11-09-2010, 06:26 PM	#6
ezza New Member Join Date: Sep 2010 Location: Melbourne Posts: 6	Originally posted by tcwilson Thanks for the information. If the camera really combines pixels via software to get the resized image that raises an interesting question about noise, since the pixels are analog sensors. I know an electrical engineer could straighten me out on this, but here's my theory. For the sake of example, we set the camera to produce a 6 mp image instead of 12mp By combining two real pixels into the new single "virtual" pixel" you would double both the signal and the noise. So the S/N ratio would remain the same. But since you have now doubled the total signal, you can attenuate it lower, also lowering the noise below the threshold of detection. Now in experimenting with the different settings, this clearly isn't the case. But if engineers could figure it out, it would be neat trick. Just imagine the low light response you'd have at a 2mp setting. Fujifilm did something along those lines in a few of their P&S cameras, called EXR. dpreview talk a bit about how it works in this review: Fujifilm FinePix F200EXR Review: 2. EXR technology: Digital Photography Review I owned the F200 for a while and I can confirm that the system does work well, although I found it more useful for giving me extra dynamic range than reducing noise.

11-10-2010, 07:29 PM	#7
tcwilson Site Supporter Join Date: Nov 2010 Location: Cincinnati, Ohio Photos: Gallery Posts: 70 Original Poster	Thank you for the information. I appreciate your taking the time to reply.