How does the sensor image become the file?

[BobShaw]

I am trying to get an understanding of the image processing in the camera, specifically for the Sony IMX161 50MP sensor used in the X1D and GFX50S if that makes a difference.

So, is the data coming from the sensor a monochromatic 14 bit value that is changed to 48 bit, (or 16 bit x 3 colour) RGB by the processor in the camera, meaning interpolation is done by the camera, or
Is the data coming out of the sensor already 48 bit (or some other) RGB meaning that interpolation is done by the sensor, or some other combination, please?

Bob

[fdisilvestro]

It depends if you are talking about the raw image or a rendered image (usually jpeg).

The data from the sensor is monochromatic 14 bit, so the raw file.

The processor in camera can interpolate to a rendered image, usually 24 bit RBG (8 bits per channel). As far as I know, most cameras don't output 16 bits per channel rendered images (RGB). Either jpeg or 8bit tiff.

The other option is to do the interpolation from the raw file out of the camera, using your choice of raw converter (LR, Phocus, C1, etc)

[BobShaw]

It depends if you are talking about the raw image or a rendered image (usually jpeg).

Raw file only

[fdisilvestro]

Raw file only

Then it is monochromatic 14 bit file, which includes a rendered jpeg preview (8 bit per channel RGB). This preview has been interpolated and rendered by the in-camera processor,

[EricV]

To be a little more explicit:

• Each pixel on the sensor records a digital number (typically 12-16 bits) representing monochromatic light intensity.
• Color is introduced by placing a color filter over each pixel.  So in that sense, the pixels do record RGB colors.

So your raw file contains 50M 16-bit pixel values, along with an RGB tag for each pixel.  In a subsequent processing step, the raw file is converted to a processed file containing 50M (3*16)-bit pixel values, where each pixel now contains all three RGB colors.

[BobShaw]

To be a little more explicit:

• Each pixel on the sensor records a digital number (typically 12-16 bits) representing monochromatic light intensity.
• Color is introduced by placing a color filter over each pixel.  So in that sense, the pixels do record RGB colors.

So your raw file contains 50M 16-bit pixel values, along with an RGB tag for each pixel.  In a subsequent processing step, the raw file is converted to a processed file containing 50M (3*16)-bit pixel values, where each pixel now contains all three RGB colors.

Thanks Eric.
So 12-16 bits of data are clocked out of the sensor chip to some buffer, where adjacent pixels are compared and interpolated outside of the sensor chip to make 48 bits per pixel. These are then saved to the raw file with additional information like lens information.
It would be valid to say then that as interpretation of colour is done by the camera and not the sensor, then different cameras using the same sensor can give different colour.

[mcbroomf]

It must be done "live" to allow us to look at the image in an EVF and on the LCD with very little lag, but as said, only on a small downsized image.  The full rez raw file is demosaiced by LR/PS/C1 etc when we do our Post.

[EricV]

https://www.imatest.com/docs/raw/