The result is still a mosaicked NEF file, i.e. demosaicking the Bayer grid CFA is still needed.
The output pixel values are computed using easy pixel-to-pixel averaging or min-max etc.
Sensor shift NEFX file output from NEF files input
For full-pixel shift, the process simply fills the proper RGB values from shifted RAW captures. The G channel being twice as dense as R and B, involves averaging more than one readings of the G channel.
When sub-pixel shift is used, the process is more complex
The result is fully demosaicked.
Google Gemini summarizes as :
The Nikon NEFX format is a merged pixel shift image format rather than a traditional RAW file. Since it is created by merging multiple images, it does not require demosaicking, which is typically needed for single-shot RAW files that use a Bayer filter. Instead, NEFX files retain RAW-like data but have already undergone pixel merging, meaning they are processed differently from standard NEF files
A raw file is a file which is (1) produced by a camera or other imaging devices, and (2) that contains the unprocessed (or minimally processed) data directly from the camera’s sensor, which is not directly suitable for viewing or printing without further processing steps such as conversion to a color profile from the native sensor color space, lens correction, denoising, and demosaicing (conversion from the camera color planes, subsampled through a Color Filter Array (CFA) or not).
Some raw files can be produced after initial capture, such as in-camera multiple exposure modes where the camera combines several exposures into a single raw file.
A raw file can combine multiple exposures, often as part of computational photography techniques (e.g., for High Dynamic Range or noise reduction).
Most current flagship phone raw files are pre-processed from multiple exposures using computational photography techniques and are often delivered as “Linear DNGs,” meaning they are already demosaiced by the phone’s image processor. They offer significant benefits such as extended quantization depth (more tonal information), reduced risk of clipping from color balancing, and a wider internal color gamut compared with standard sRGB or Adobe RGB.
Foveon RAW files do not need demosaicing because their unique sensor design captures full spectral information at each pixel location by utilizing stacked photodiodes with different depth-dependent light absorption characteristics, unlike sensors with a Color Filter Array. While these layers are broadly sensitive to what we colloquially call “blue,” “green,” and “red” light, the raw output is in the sensor’s native color space and requires a color transformation matrix (rather than interpolation) to be converted into a standard RGB color space.
At this current time, DxO PhotoLab functions as a converter for both traditional raw files (which it demosaics) and pre-demosaiced files (like “Linear DNGs” from certain phones). For example, the Apple iPhone 8 DNG format, which is typically already demosaiced to some extent, is supported by DxO PhotoLab.
NEF (Nikon Electronic Format) files are Nikon’s proprietary raw format, containing the raw, Bayer-patterned sensor data that requires demosaicing by a raw converter to produce a full-color image.
However, NEFX files, which are the output of Nikon’s in-camera or NX Studio pixel-shift merging process, are a special case. Due to the pixel-shifting technique, where multiple individual exposures are combined to capture full color information for each pixel location, the NEFX file is effectively a “3-color image” (RGB data per pixel) and therefore does not require further demosaicing by a raw converter. At present, DxO PhotoLab does not natively support Nikon NEFX files; users typically need to convert them to a compatible format (like TIFF) using Nikon’s NX Studio before importing into DxO PhotoLab.
Other widely used RAW converters that accept Nikon NEFX include Nikon NX Studio, Adobe Photoshop (via Camera Raw), Adobe Bridge, Adobe DNG Converter, and Capture One.
Thanks for raising this support for Nikon’s HE\ (NEFX) files would be a great addition. With more cameras adopting this format, having native processing in DxO PhotoLab would really streamline workflows for Nikon users. Hope this gets on the roadmap soon!
That’s what OM System’s Hi-Res mode does, with the result being a RAW/ORF file (plus the first shot taken, of the 8 or 12 image set, for comparison … also saved as a RAW/ORI file).
The way Nikon implements Pixel-Shift can be either to :-
Take repeated images with no sensor movement.
Take a complete set of images with the sensor moving 1/2 pixel in all 3 other directions.
or a mixture of both.
I’m guessing 1) does catch those values?
Although, it makes sense to ‘do’ the full 32 frame blend/stack if one is going to the bother of setting up for this type of photography.
A sensel has an analogue output value which is being digitized.
But my remark was what is the meaning of taking repeated images without sensor movement in this thread?
AFAIK, a sensel is the individual light sensitive cell under each of the 4 colour filter elements, ie RGBG to make an individual pixel. So 4 sensels to one pixel.
There are people who remove the colour filter layer (CFA) to make a high res mono sensor… or then there’s the mono astro sensors with no CFA layer.
If you ask me personal, no. If you ask in general, yes.
I never heard of this way of noise reduction and i don’t think you reduce noise then. I only know of taking a second image with the same time but without lighting the sensor and subtracking the values.
Sensor noise is random.
If you take multiple images and average (or maybe median?) them mathamatically, the noise is greatly reduced.
Whether this is at the sensel stage or the pixel stage, the result is the same.
