Bayer interpolation

[RalphLambrecht]

Sorry if this was discussed before but I'm confused by different sources making different claims of how much resolution the Bayer interpolation is costing a digital camera.Some forces claim camera resolution has to be divided by 3 ;other claim it has to be divided by 4.Which is it and Is there any loss in resolution if the cameras set to capture monochrome?
Thanks in advance

 

[Chan Tran]

Sorry if this was discussed before but I'm confused by different sources making different claims of how much resolution the Bayer interpolation is costing a digital camera.Some forces claim camera resolution has to be divided by 3 ;other claim it has to be divided by 4.Which is it and Is there any loss in resolution if the cameras set to capture monochrome?
Thanks in advance

Bayer passed away in 2012 otherwise may be he can tell.

 

[RalphLambrecht]

Bayer passed away in 2012 otherwise may be he can tell.

that's too bad.I'm starting to think,there may be no loss of resolution at all and all is regained through some fancy software interpolation;after all,I'm still getting an image 7390px by 4912px even in color.

 

[rbultman]

I think the short answer is that there is no "divide by" and you have the full pixel dimensions in the resulting image.

I think the longer answer is more complex and involves considerations of pixel dimensions, color accuracy, and the ability to resolve fine details.

Each photo site in a classic Bayer sensor only senses one color, R G or B. In order to arrive at an RGB value for each pixel in the resulting image, a demosaicing algorithm is applied to interpolate the color using the adjacent pixels. So in terms of pixel count, you don't lose any pixels. But, interpolation can lead to false color and other artifacts which are well known and well covered. The Foveon sensor in theory does not suffer from this since each photo site records the full spectrum via 3 different, stacked photo diodes. It suffers from its own problems but at least as covered by DP Review in April 10, 2012, the Foveon sensor is capable of "exceptional image quality at low ISO." I have no idea how the image quality compares to new Bayer-type sensors.

I think the "divide by" that you mention is most likely 4 since per the Bayer pattern, there are twice as many green sites as there are red or blue. If you look at a 2x2 section of a classic Bayer array, it looks something like this:

RG
GB

But there is not really a "divide by" since each spectrum-limited photo site is treated as a full-color photo site via the demosaic algorithm.

Regards,
Rob

 

[RalphLambrecht]

I think the short answer is that there is no "divide by" and you have the full pixel dimensions in the resulting image.

I think the longer answer is more complex and involves considerations of pixel dimensions, color accuracy, and the ability to resolve fine details.

Each photo site in a classic Bayer sensor only senses one color, R G or B. In order to arrive at an RGB value for each pixel in the resulting image, a demosaicing algorithm is applied to interpolate the color using the adjacent pixels. So in terms of pixel count, you don't lose any pixels. But, interpolation can lead to false color and other artifacts which are well known and well covered. The Foveon sensor in theory does not suffer from this since each photo site records the full spectrum via 3 different, stacked photo diodes. It suffers from its own problems but at least as covered by DP Review in April 10, 2012, the Foveon sensor is capable of "exceptional image quality at low ISO." I have no idea how the image quality compares to new Bayer-type sensors.

I think the "divide by" that you mention is most likely 4 since per the Bayer pattern, there are twice as many green sites as there are red or blue. If you look at a 2x2 section of a classic Bayer array, it looks something like this:

RG
GB

But there is not really a "divide by" since each spectrum-limited photo site is treated as a full-color photo site via the demosaic algorithm.

Regards,
Rob

I came to the same conclusion but msny thanks for the effort to explain this to me.Thr Bayer filter turns an essential monochrome sensor into a color capturing device without losing resolution;great

 

[rbultman]

You don't lose resolution in the sense of pixel count but may lose it in the sense of resolving power as in lp/mm. False color effects affect this. See the resolution charts for your favorite digital camera on DP Review. That sort of "loss" may be completely immaterial to you or not depending on what you photograph and how your images are displayed.

Regards,
Rob

Sent from my Nexus 5 using Tapatalk

 

[RalphLambrecht]

You don't lose resolution in the sense of pixel count but may lose it in the sense of resolving power as in lp/mm. False color effects affect this. See the resolution charts for your favorite digital camera on DP Review. That sort of "loss" may be completely immaterial to you or not depending on what you photograph and how your images are displayed.

Regards,
Rob

Sent from my Nexus 5 using Tapatalk

true but one step at a time

 

[rbultman]

What do the rows and columns mean?

Sent from my Nexus 5 using Tapatalk

 

[RalphLambrecht]

What do the rows and columns mean?

Sent from my Nexus 5 using Tapatalk

Sorry, it's not more self evident;top row shows film resolution and bottom row the Mpx required to match it in digital.the assumption is that there is no loss due to the Bayer filter,which may or may not be true.

 

[Chan Tran]

You don't lose any pixels as the number of pixels are the same and after the conversion from RAW each pixel has the full RGB values instead of each pixel only has either R or G or B value. But that actual amount of details I believe is less than that number of pixels could convey but significantly better than the case where you simply use 3 single color pixel to make on full color pixel. I believe the algorithm looks at the adjacent pixels and figure out what's the most likely values are.