X-Trans Performance as Shown with DPR Studio Shot Comparison Tool

[toktik]

I was evaluating various cameras using the DPR Studio Shot Comparison Tool and I noticed the raw output for Fujifilm X-Trans cameras show better results related to false colors than many other systems, including GFX cameras. I am refering to the color patterns that are visible in the drawings on the left and right side. Here is an example:



Is this proof that X-Trans sensors are much better at avoiding issues such as moire, aliasing, etc. than cameras that use a Bayer sensor, or am I misunderstanding what I see?

 

[Powerdoc]

I was evaluating various cameras using the DPR Studio Shot Comparison Tool and I noticed the raw output for Fujifilm X-Trans cameras show better results related to false colors than many other systems, including GFX cameras. I am refering to the color patterns that are visible in the drawings on the left and right side. Here is an example:



Is this proof that X-Trans sensors are much better at avoiding issues such as moire, aliasing, etc. than cameras that use a Bayer sensor, or am I misunderstanding what I see?

No it's normal , X trans have this advantage. It also have backside effect.

--
It's all about photography

 

[toktik]

I was evaluating various cameras using the DPR Studio Shot Comparison Tool and I noticed the raw output for Fujifilm X-Trans cameras show better results related to false colors than many other systems, including GFX cameras. I am refering to the color patterns that are visible in the drawings on the left and right side. Here is an example:



Is this proof that X-Trans sensors are much better at avoiding issues such as moire, aliasing, etc. than cameras that use a Bayer sensor, or am I misunderstanding what I see?

No it's normal , X trans have this advantage. It also have backside effect.

Thank you for responding. What do you mean by backside effect?

 

[SigmaChrome]

This is a perfect example of why Fuji MF cameras should be using X-Trans sensors -- or even Foveon...

 

[toktik]

This is a perfect example of why Fuji MF cameras should be using X-Trans sensors -- or even Foveon...

The GFX 100s shows very little of those color issues, but the lower resolution sensor certainly does not perform as well in that area. My understanding is that X-Trans sensors are not used for GFX because the time required to store an image would increase substantially. I assume the processing in much more complex than if a Bayer sensor is used.

 

[Erik Baumgartner]

I was evaluating various cameras using the DPR Studio Shot Comparison Tool and I noticed the raw output for Fujifilm X-Trans cameras show better results related to false colors than many other systems, including GFX cameras. I am refering to the color patterns that are visible in the drawings on the left and right side. Here is an example:



Is this proof that X-Trans sensors are much better at avoiding issues such as moire, aliasing, etc. than cameras that use a Bayer sensor, or am I misunderstanding what I see?

The GFX shows more of a moire issue because it's of it's relatively low resolution at a much larger sensor size and lack of an AA filter, moire tends to be much less of an issue as the resolution increases. X-Trans isn't immune to moire either, and when it does happen. it's usually in different and more more difficult to remove places. With the 24/26MB APS-C sensors, any slight moire advantages with X-Trans are outweighed (IMO) by its demosaicing issues (which produce false detail rather than false color). Personally, I'd rather the new cameras be Bayer, but I'm used to X-Trans and can deal with it either way.


GFX 50R here. Unlike some X-Trans moire, Bayer moire is usually pretty easy to deal with in post when it occurs (with Lightroom, anyway,

 

[toktik]

I was evaluating various cameras using the DPR Studio Shot Comparison Tool and I noticed the raw output for Fujifilm X-Trans cameras show better results related to false colors than many other systems, including GFX cameras. I am refering to the color patterns that are visible in the drawings on the left and right side. Here is an example:



Is this proof that X-Trans sensors are much better at avoiding issues such as moire, aliasing, etc. than cameras that use a Bayer sensor, or am I misunderstanding what I see?

The GFX shows more of a moire issue because it's of it's relatively low resolution at a much larger sensor size and lack of an AA filter, moire tends to be much less of an issue as the resolution increases. X-Trans isn't immune to moire either, and when it does happen. it's usually in different and more more difficult to remove places. With the 24/26MB APS-C sensors, any slight moire advantages with X-Trans are outweighed (IMO) by its demosaicing issues (which produce false detail rather than false color). Personally, I'd rather the new cameras be Bayer, but I'm used to X-Trans and can deal with it either way.


GFX 50R here. Unlike some X-Trans moire, Bayer moire is usually pretty easy to deal with in post when it occurs (with Lightroom, anyway,

Thank you for the information. I am curious how the 40MP X-Trans sensor will perform. Hopefully the X-H2 will be added to the DPR tool soon.

 

[Michael Everett]

I was evaluating various cameras using the DPR Studio Shot Comparison Tool and I noticed the raw output for Fujifilm X-Trans cameras show better results related to false colors than many other systems, including GFX cameras. I am refering to the color patterns that are visible in the drawings on the left and right side. Here is an example:



Is this proof that X-Trans sensors are much better at avoiding issues such as moire, aliasing, etc. than cameras that use a Bayer sensor, or am I misunderstanding what I see?

No it's normal , X trans have this advantage. It also have backside effect.

Thank you for responding. What do you mean by backside effect?

I am curious too as to what backside effect means.

 

[Erik Baumgartner]

I was evaluating various cameras using the DPR Studio Shot Comparison Tool and I noticed the raw output for Fujifilm X-Trans cameras show better results related to false colors than many other systems, including GFX cameras. I am refering to the color patterns that are visible in the drawings on the left and right side. Here is an example:



Is this proof that X-Trans sensors are much better at avoiding issues such as moire, aliasing, etc. than cameras that use a Bayer sensor, or am I misunderstanding what I see?

The GFX shows more of a moire issue because it's of it's relatively low resolution at a much larger sensor size and lack of an AA filter, moire tends to be much less of an issue as the resolution increases. X-Trans isn't immune to moire either, and when it does happen. it's usually in different and more more difficult to remove places. With the 24/26MB APS-C sensors, any slight moire advantages with X-Trans are outweighed (IMO) by its demosaicing issues (which produce false detail rather than false color). Personally, I'd rather the new cameras be Bayer, but I'm used to X-Trans and can deal with it either way.


GFX 50R here. Unlike some X-Trans moire, Bayer moire is usually pretty easy to deal with in post when it occurs (with Lightroom, anyway,

Thank you for the information. I am curious how the 40MP X-Trans sensor will perform. Hopefully the X-H2 will be added to the DPR tool soon.

The higher resolution should greatly reduce any X-Trans demosaicing errors (improving the image quality with all lenses), so I'm looking forward to maybe going for an X-T5 when it finally comes along. I'm hoping they managed to keep the high ISO noise on par with the older cameras.

 

[Truman Prevatt]

XTrans was designed to mitigate one effect for aliasing. The XTrans not mitigate aliasing. It mitigates the color Moire that results from the subsampling of a Bayer pattern. The aliasing is still present it, however, does not show up as the false color.

Many find the false color Moire very objectionable.

 

[Powerdoc]

I was evaluating various cameras using the DPR Studio Shot Comparison Tool and I noticed the raw output for Fujifilm X-Trans cameras show better results related to false colors than many other systems, including GFX cameras. I am refering to the color patterns that are visible in the drawings on the left and right side. Here is an example:



Is this proof that X-Trans sensors are much better at avoiding issues such as moire, aliasing, etc. than cameras that use a Bayer sensor, or am I misunderstanding what I see?

No it's normal , X trans have this advantage. It also have backside effect.

Thank you for responding. What do you mean by backside effect?

The X trans matrix is big (6 by 6) and it can lead in demosaicing issues. It reduce chroma noise but it also reduce color information.



--
It's all about photography

 

[saltydogstudios]

I was evaluating various cameras using the DPR Studio Shot Comparison Tool and I noticed the raw output for Fujifilm X-Trans cameras show better results related to false colors than many other systems, including GFX cameras. I am refering to the color patterns that are visible in the drawings on the left and right side. Here is an example:



Is this proof that X-Trans sensors are much better at avoiding issues such as moire, aliasing, etc. than cameras that use a Bayer sensor, or am I misunderstanding what I see?

No it's normal , X trans have this advantage. It also have backside effect.

Thank you for responding. What do you mean by backside effect?

The X trans matrix is big (6 by 6) and it can lead in demosaicing issues. It reduce chroma noise but it also reduce color information.

Expanding on what Powerdoc said and providing some background information that I think the OP could use:

Moire is what you get when you have two regular grids one laid atop the other. The grids combine in weird patterns, producing false colors. This isn't limited to photography. Think of a window screen - it's a regular grid. If you doubled up the grid, any slight variance between the grids will produce a pattern that causes "false colors" or patterns to appear as each of the two grid patterns interfere with each other.

Once you know what this looks like, you'll start to see it more often in the world of man made materials with regular grids.

(most) Cameras require each pixel to have a color filter atop in order for it to "see" color. Red or Green or Blue. This is known as a Color Filter Array (CFA).

Bayer pattern sensors arrange the CFA in a regular 2x2 grid and is prone to Moire. Two green, one red, one blue. (two green because the eye is more sensitive to green)

This mainly shows up when photographing clothing. To get around this, camera manufacturers purposefully put a blur atop the sensor (Anti Aliasing or AA filter). This produced slightly less sharp images that were less prone to Moire.

Starting with the Nikon D810, camera manufacturers introduced more sensors without the AA filter. This increased sharpness (less blur) but made cameras more prone to moire.

The X-Trans sensor was designed to mitigate this by having a "more random" sensor - a 6x6 grid such that the pixels. 2x2 green, and blue and red pixels surrounding it.

In bother Bayer and X-Trans, each pixel is either red or green or blue, and at least one neighbor is each of the other colors.

Each Bayer pixel is surrounded by other colors - giving the camera a lot of information to reconstruct a color image.

Each X-Trans sensor may have just 1 neighboring pixel that's another color, giving it less information to reconstruct a color image.

Hence reduced color information.

Which you choose - Bayer with AA filter, Bayer with no AA filter, X-Trans - is a choice each of us make and the compromises are (at least partially) determined by what we choose to photograph or prioritize in our photography.

Photograph nature and want sharp images? Bayer with no AA filter.

Photograph clothing (and maybe architecture?) and shoot in color? Bayer with AA filter or X-Trans.

Everything else sort of falls in between those two extremes.



--
"no one should have a camera that can't play Candy Crush Saga."
"I've been saying this for years. There is a difference between people who buy gear and those who use it." - https://www.dpreview.com/forums/post/65815232
Camera JPG Portrait Shootout: https://www.dpreview.com/forums/thread/4492044
Great Cinematography: https://www.dpreview.com/forums/thread/4498434
Blog: http://sodium.nyc/blog/
Sometimes I take photos: https://www.instagram.com/sodiumstudio/

 

[Truman Prevatt]

This is a perfect example of why Fuji MF cameras should be using X-Trans sensors -- or even Foveon...

The GFX 100s shows very little of those color issues, but the lower resolution sensor certainly does not perform as well in that area. My understanding is that X-Trans sensors are not used for GFX because the time required to store an image would increase substantially. I assume the processing in much more complex than if a Bayer sensor is used.

Aliasing is a phenomenon of digital sampling. It doesn't matter if the data being sampled is audio for CD or streaming digital music, microwave RF off an antenna for cell phone service or fiber optics or satellite communications - aliasing is a mathematical reality of digital sampling. If you take a sine wave with a frequency of two cycles and sample it two times a sample you will not get a sine wave - you will get a constant since the sampling happens at the same value of the sine wave! We cannot reconstruct the original signal from the samples.

Here is a pretty good tutorial on what happens.

Mathematically it results from the fact that the space of all functions whose Fourier transform vanish outside a fixed compact interval is isomorphic to the set of sequences on the integers.

if we sample at more than two times a period - we can perfectly reconstruct the signal. This is what the Shannon-Nyquist sampling theorem from information theory tells us. It doesn't matter if where the sine wave comes from - be it out of a guitar or an optical line pair. One sine wave (guitar) is indexed by time and the line pair is interred by space.

Aliasing occurs when the sampling is less than twice per cycle. We can reconstruct a sine wave from the samples. However, it is not the correct sine wave - it's frequency is different. It will give a frequency at the input frequency - Shannon-Nyquist sampling frequency. In an image it will give detail but false detail at a lower spatial frequency.

Aliasing occurs in all digital sensors. It occurs in my Leica Q2M which is an black and white sensor. However, things get sketchy when we demand color images from sensors by putting a color filter array over the sensor. The Bayer array results in 3 sensors, one green, one blue and one red with periodic sampling sensors. The problem is the red and blue sensors are sampled at half the rate of the green. This results in color artifacts known as color Moire.

While the XTrans does not in general produce the classic color Moire on high frequency patterns, it does produce aliasing as all sampling systems do. The aliased energy is simply spread over the multiple frequencies over the image and appears in the noise. With a Bayer it appears as single frequencies on the image.

Every thing is a trade off. The key is more MP to raise the Shannon-Nyquist sampling frequency.