The X20's Green Problem

Started May 3, 2013 | Discussions
Nukunukoo
Regular MemberPosts: 353Gear list
Like?
The X20's Green Problem
May 3, 2013

After reading DPR’s excellent review of the X20 some of the problems they pointed out grabbed my attention. Specifically, the way the X20 (or probably the X-Trans in general) captures green subjects. While most small sensor compacts smudge images at low ISOs, the X20 smudges green details more than expected.

This brings me to the suspicion that the X-Trans sensor could be the main cause of this.

If you look at the X-Trans color array, it covers a 6x6 pattern as opposed to the Bayer’s 2x2. Furthermore, smack right in the middle of the X-Trans matrix is a solid block of 2x2 green. This relatively huge area does pose a demosaicing challenge.

The X-Trans 6x6 pattern with a 2x2 green block in the middle.

I wanted to do a quick test to see if the X20 does treat anything green differently from the rest of the colors, namely, red and blue. Still on a holiday, all I have at my disposal (besides my X20) are my Mac Air and Samsung 7.7 tablet. I created three single colored spiral images for each of the RGB colors, where each individual RGB color is given its max value (255/FF) and assigning a zero to the other two colors for each image.

Since my Mac was a non-retina version, I opted to capture the images on my Tab 7.7, which has a respectable DPI of 196 and is currently the biggest SAMOLED display in production so I have no doubt with regards to its contrast and saturation. I seated the X20 on top of a book (my Gorillapod is with a friend at the moment). All things considered, this is as scientific as I can get!

The Samsung Tab 7.7 testbed.

I also included a black spiral for reference (now that I think about it, a 50% gray would have been better!). Below are the four color swatches cropped and increased by 400 using the Nearest Neighbor method to eliminate any gradient interpolation. The images were taken from RAW to remove any JPEG-engine related color bias and opened in LR5. LR5 is only used to open the captured images and no post-processing was used including NR and sharpening.

Green (400%)

Blue (400%)

Black (400%)

Below are the images side-by-side in both color and grayscale.

The images next to each other

In grayscale

Notice the pronounced vertical artefact on the green spiral. This artefact is also visible to a lesser degree on the blue image, which is not surprising since blue is a component of green. It’s hard to conclude that the X-Trans is to blame: conducting a similar test with all of the X-Trans Fuji is needed. What this experiment do show is that the X20 images exhibit some green and blue artefacting.

In all fairness, with my real world tests (showing my works, both printed and onscreen) to fellow colleagues, they did not notice anything off with my JPEG images until I asked them to pixel-peep on specific areas that was most affected by the artefacting. Even then, they needed to be told that something was off.

Still, this does not mean all is forgiven. Whether a firmware upgrade is a solution or that the X-Trans sensor’s big green “hole” has an inherent flaw, Fuji needs to address this issue to clear the air for current and future X20 users.

Liz.

 Nukunukoo's gear list:Nukunukoo's gear list
Fujifilm X20 Nikon D300S Nikon D800E Nikon D7100 Sigma 50-150mm F2.8 EX DC APO OS HSM +4 more
Fujifilm X20
If you believe there are incorrect tags, please send us this post using our feedback form.
Photonics
Regular MemberPosts: 120
Like?
Re: The X20's Green Problem
In reply to Nukunukoo, May 3, 2013

thanks for the research  

Reply   Reply with quote   Complain
Gaze
Contributing MemberPosts: 627
Like?
Re: The X20's Green Problem
In reply to Nukunukoo, May 3, 2013

Hi Liz, you said that "Fuji needs to address this issue ….." but did you try to contact Fuji about this issue and if so, what was their response? …. just curious …

Reply   Reply with quote   Complain
Nukunukoo
Regular MemberPosts: 353Gear list
Like?
Re: The X20's Green Problem
In reply to Gaze, May 3, 2013

Gaze wrote:

Hi Liz, you said that "Fuji needs to address this issue ….." but did you try to contact Fuji about this issue and if so, what was their response? …. just curious …

Not yet. But I will forward this concern soon. Later, I'll be posting some full sized landscape shots (lotsa greens) to see how much of an issue this oddity really is. In my opinion, not much. But like I said, it's something I can't just ignore as well.

Liz

 Nukunukoo's gear list:Nukunukoo's gear list
Fujifilm X20 Nikon D300S Nikon D800E Nikon D7100 Sigma 50-150mm F2.8 EX DC APO OS HSM +4 more
Reply   Reply with quote   Complain
XavierP56
Contributing MemberPosts: 919
Like?
Re: The X20's Green Problem
In reply to Nukunukoo, May 3, 2013

In real world, I see nothing wrong with the X20 green.

Shooting RAW, LR 4.4 / 5.

Very happy with X20.

Reply   Reply with quote   Complain
CAcreeks
Senior MemberPosts: 6,434
Like?
Great post, Liz!
In reply to Nukunukoo, May 3, 2013

Something about the X20's greens bothered me from the beginning, but I could not put my finger on why. You did.

Green foliage etc. did not seem smeared, it just seemed too green, or artificial, but I guess smearing made it seem that way.

The black specs visible when pixel peeping do not bother me as much.

Sincerely, still an EXR fanboy.

Reply   Reply with quote   Complain
Dorus
Senior MemberPosts: 1,065
Like?
Re: The X20's Green Problem - Brilliant test
In reply to Nukunukoo, May 3, 2013

Pretty clever and well done! You have clearly demonstrated a difference between how green is rendered vs. the other colors, and you did that based on an original observation and sound hypothesis. Whether or not that translates to a problem in real-life pictures is almost secondary, and will differ between users.

Dorus

Reply   Reply with quote   Complain
Nukunukoo
Regular MemberPosts: 353Gear list
Like?
Re: Great post, Liz!
In reply to CAcreeks, May 3, 2013

CAcreeks wrote:

Something about the X20's greens bothered me from the beginning, but I could not put my finger on why. You did.

Green foliage etc. did not seem smeared, it just seemed too green, or artificial, but I guess smearing made it seem that way.

The black specs visible when pixel peeping do not bother me as much.

Sincerely, still an EXR fanboy.

Indeed. Probably, if Fuji stuck to the OLPF-less Bayer pattern and since the sensor is BSI to begin with, they would not have this smear/RAW incompatibility problems.

Liz.

 Nukunukoo's gear list:Nukunukoo's gear list
Fujifilm X20 Nikon D300S Nikon D800E Nikon D7100 Sigma 50-150mm F2.8 EX DC APO OS HSM +4 more
Reply   Reply with quote   Complain
CAcreeks
Senior MemberPosts: 6,434
Like?
Bayer sensor was NIH
In reply to Nukunukoo, May 3, 2013

Nukunukoo wrote:

Indeed. Probably, if Fuji stuck to the OLPF-less Bayer pattern and since the sensor is BSI to begin with, they would not have this smear/RAW incompatibility problems.

The Bayer pattern sensor was invented by some guy (forgot his name, maybe Bruce Bayer?) working at Kodak.

Not Invented Here!

Seems like every new color filter array that Fujifilm comes up with is worse than the last. SuperCCD SR has die-hard fans, but I never used it. For dynamic range EXR trades off resolution, which is not very important to me as a website author.

I dunno, have not made up my mind, maybe X-Trans is good for larger sensors.

P.S. Probably Fujifilm abandoned SuperCCD because of its non-standard hexagonal sensor pattern, which was expensive to manufacture in small quantities. EXR is basically a normal (e.g. Sony) sensor rotated 45 degrees. Most are made by Toshiba now. X-Trans is a regular sensor, unrotated, with a different color filter on top. Cheap to make!

Reply   Reply with quote   Complain
Nukunukoo
Regular MemberPosts: 353Gear list
Like?
Re: The X20's Green Problem - Brilliant test
In reply to Dorus, May 3, 2013

Dorus wrote:

Pretty clever and well done! You have clearly demonstrated a difference between how green is rendered vs. the other colors, and you did that based on an original observation and sound hypothesis. Whether or not that translates to a problem in real-life pictures is almost secondary, and will differ between users.

Dorus

Thanks Dorus!

In most cases, the issue never crop up in my real-world applications. Take the photo below, for instance: (JPEG, mid-compression)

You won't notice anything unless you pixel-peep. I also have some grass photos of low contrasts that has more smudging but barely noticeable (unless you pixel-peep as a profession!)

Liz

 Nukunukoo's gear list:Nukunukoo's gear list
Fujifilm X20 Nikon D300S Nikon D800E Nikon D7100 Sigma 50-150mm F2.8 EX DC APO OS HSM +4 more
Reply   Reply with quote   Complain
Leo33
Senior MemberPosts: 2,686Gear list
Like?
Re: The X20's Green Problem
In reply to Nukunukoo, May 3, 2013

Excellent detective work, Liz. Besides the sensor, I feel Fuji needs to beef up the Jpeg engine for the X-20 as well.  :)))
Leo33

 Leo33's gear list:Leo33's gear list
Fujifilm XF1
Reply   Reply with quote   Complain
Nukunukoo
Regular MemberPosts: 353Gear list
Like?
Re: The X20's Green Problem
In reply to Leo33, May 4, 2013

Leo33 wrote:

Excellent detective work, Liz. Besides the sensor, I feel Fuji needs to beef up the Jpeg engine for the X-20 as well.  :)))
Leo33

Or perhaps use a different demosaicing algorithm altogether!

Liz.

 Nukunukoo's gear list:Nukunukoo's gear list
Fujifilm X20 Nikon D300S Nikon D800E Nikon D7100 Sigma 50-150mm F2.8 EX DC APO OS HSM +4 more
Reply   Reply with quote   Complain
tomhongkong
Senior MemberPosts: 2,441
Like?
Re: The X20's Green Problem
In reply to Nukunukoo, May 4, 2013

I agree, a very interesting piece of work.  However how relevent it is unless you pixel peep to an extreme extent is worthwhile asking.

Let's face it, most users of X20 (or X10) are not likely to print at greater than 50% and both cameras produce excellent results at this enlarhgement level.  Screen viewing is irrelevent unless you are pixel peeping as 1920x1080 will not show the problem.

I would hate to have the whole X20 scene driven by some academic view of images at 100% (or even larger) when it has not real significance for the bulk of users.

Not to diminish your work, which is interesting

tom

Reply   Reply with quote   Complain
Nukunukoo
Regular MemberPosts: 353Gear list
Like?
Re: The X20's Green Problem
In reply to tomhongkong, May 4, 2013

tomhongkong wrote:

I agree, a very interesting piece of work.  However how relevent it is unless you pixel peep to an extreme extent is worthwhile asking.

Let's face it, most users of X20 (or X10) are not likely to print at greater than 50% and both cameras produce excellent results at this enlarhgement level.  Screen viewing is irrelevent unless you are pixel peeping as 1920x1080 will not show the problem.

I would hate to have the whole X20 scene driven by some academic view of images at 100% (or even larger) when it has not real significance for the bulk of users.

Not to diminish your work, which is interesting

tom

Indeed, however, there are people whose life seem to revolve around the hobby of pixel peeping and losing the joys of actually appreciating the whole composition!

Liz.

 Nukunukoo's gear list:Nukunukoo's gear list
Fujifilm X20 Nikon D300S Nikon D800E Nikon D7100 Sigma 50-150mm F2.8 EX DC APO OS HSM +4 more
Reply   Reply with quote   Complain
steras
Contributing MemberPosts: 770
Like?
Liz, this is a clever test...
In reply to Nukunukoo, May 4, 2013

which I managed to reproduce. However, when I tilted the camera 30-40 degrees relative to the target the vertical pattern in the green target disappeared. Any thoughts?

Reply   Reply with quote   Complain
yellodog
Senior MemberPosts: 1,075Gear list
Like?
Re: The X20's Green Problem
In reply to Nukunukoo, May 4, 2013

Brilliant work. Something Fuji must have done at some point but decided it was a reasonable trade off for the good detail? Or in too much of a hurry to address the problem although it seems the four color map theorem might have been useful?

Reply   Reply with quote   Complain
brianj
Forum ProPosts: 13,327Gear list
Like?
Re: The X20's Green Problem
In reply to Nukunukoo, May 4, 2013

Why dwell on this minor imperfection when the entire bayer system is riddled with flaws and this is just a derivitive of that.  We expect this outdated method for sensors to be a compromise, so most people just ignore it.

The foveron principal gets closer the something useful but even it has many problems.  Its good that fuji is actually attempting to find a better sensor layout, but I think we are a long way from the ultimate solution.

So as some have said, best to get on with photography.

Brian

 brianj's gear list:brianj's gear list
Canon PowerShot ELPH 330 HS
Reply   Reply with quote   Complain
Nukunukoo
Regular MemberPosts: 353Gear list
Like?
Re: Liz, this is a clever test...
In reply to steras, May 4, 2013

steras wrote:

which I managed to reproduce. However, when I tilted the camera 30-40 degrees relative to the target the vertical pattern in the green target disappeared. Any thoughts?

If your target is an LCD/TFT backlit display, there will be a angle where the colors and images from the display will merge and eventually the lower contrast will affect the results. I was using an AMOLED screen where each pixel actually transmit light instead of being translucent behind a backlight. This has nothing to do with the X20.

Liz.

 Nukunukoo's gear list:Nukunukoo's gear list
Fujifilm X20 Nikon D300S Nikon D800E Nikon D7100 Sigma 50-150mm F2.8 EX DC APO OS HSM +4 more
Reply   Reply with quote   Complain
Nukunukoo
Regular MemberPosts: 353Gear list
Like?
Re: The X20's Green Problem
In reply to yellodog, May 4, 2013

It's also possible, however unlikely, that Adobe's X-Trans RAW converter is the culprit. I can have have the RAW files available if anyone is willing to convert  them from SilkyPix for comparison.

Liz.

 Nukunukoo's gear list:Nukunukoo's gear list
Fujifilm X20 Nikon D300S Nikon D800E Nikon D7100 Sigma 50-150mm F2.8 EX DC APO OS HSM +4 more
Reply   Reply with quote   Complain
Nukunukoo
Regular MemberPosts: 353Gear list
Like?
Re: The X20's Green Problem
In reply to brianj, May 4, 2013

brianj wrote:

Why dwell on this minor imperfection when the entire bayer system is riddled with flaws and this is just a derivitive of that.  We expect this outdated method for sensors to be a compromise, so most people just ignore it.

The foveron principal gets closer the something useful but even it has many problems.  Its good that fuji is actually attempting to find a better sensor layout, but I think we are a long way from the ultimate solution.

So as some have said, best to get on with photography.

Brian

Indeed. Cameras are just tools and ultimately, we are the photographers! While a better camera does help in making a better picture, no amount of camera will save you if you are a bad photographer to begin with.

My X20 is one fun camera and I really do enjoy using it for my casual shots! My treatment on the X-Trans is to satisfy my curiosity.

Liz.

 Nukunukoo's gear list:Nukunukoo's gear list
Fujifilm X20 Nikon D300S Nikon D800E Nikon D7100 Sigma 50-150mm F2.8 EX DC APO OS HSM +4 more
Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum MMy threads