The new Fuji X-Trans sensor in the X-Pro 1 should reduce/eliminate moiré

Original image from the movie "Friday the 13th"

Back in October 2011, Fujifilm representative claimed that their upcoming mirrorless interchangeable lens camera (X-Pro 1) will have resolution and low noise that will rival the performance of 35mm full size sensor.

Fuji's secret will be their new Fuji X-Trans sensor and EXR Pro processor. I am not very clear on the details, but the story goes like that:

Instead of the traditional 2 x 2 sensor layout where you have two equally spaced green and blue pixels, the new Fuji X-Trans sensor will have a 6 x 6 design that consists of four 3 x 3 pixels that are mixed (not arranged like in the traditional design). This design should eliminate moiré (the sensor does not have AA filter). From there the processor will form colors which will add extra CPU time, where the new EXR Processor Pro come in play.

We we know more on Monday.

Yes, Fuji designs their own sensors, cameras and lenses and other manufacturers should be afraid of Fujifilm.

This entry was posted in Fuji and tagged . Bookmark the permalink. Trackbacks are closed, but you can post a comment.
  • Better than 35mm full size sensor?…. (it must be a 70mm sensor)
    Can’t wait to see…. I hope, whish and would really like to see it coming to reallity but i guess i’m going to have to wait to find that out….

  • ozbaz

    The sad thing is Fuji is everything Kodak could have been. Kodak made sensors ; they even invented the digital camera!

    • Keep in mind that two years ago Fuji did not have any breathtaking digital products. They were able to find their niche successfully. I can see them becoming a major player in the industry.

      • Frosti7

        Actually Fuji came out with the f30 in 2006, it’s sensors low light performance is still breathtaking today

      • photonut

        They need to work on the design (so does Olympus). That retro design looked cool for a while, but is becoming outdated.

        • Twinkle

          I don’t think so, Fuji knows exactly what they’re doing in offering this design. It can’t become outdated: it’s already a classic.

      • Fuji had breathtaking digital products since 1999.

  • Neely Fallon

    R.I.P. Kodak
    Hope Leica put a call into Fuji on the REAL NEXt gen sensor for their EVIL & M10 cause Kodak CCD is gone baby, gone.

  • grayscale

    leave it to fuji to come up with the a solution for moire..
    Why didn’t anyone do this before?
    Of course, since moire is the resulting interference pattern between real world patterns and sensors colour pixel patterns, if you randomize the colour pixels there will be no interference… brilliant

    I guess they will also randomize the intensity pixels to prevent AA, the interference between continuous lines at non-normal angles.

    hmm.. it’d probably work best if they went from a grid to a honeycomb, or something with more random placement.

    • J-Man

      I don’t think moire will be 100% eliminated, just reduced, the pixel pattern(rows & columns) is still there.
      But other demosaicing artifacts may be reduced, or at least looking more film like.

    • Chris

      “leave it to fuji…” Actually, i believe it was kodak who invented and pattented this sort of solution.

      But yes, it kind of makes sense that a film company would be savy about ways to form useful bits of color information.

    • Of course, since moire is the resulting interference pattern between real world patterns and sensors colour pixel patterns, if you randomize the colour pixels there will be no interference… brilliant

      I don’t think you quite understand. Randomising will blur things.

      And no one said anything about randomising anything… unless I’ve missed it somewhere.

      Sure, this design might reduce “or eliminate” moire but let’s hope it has no impact on the detail. After all, moire is not the most detrimental thing to IQ.

      • I just noticed where all this randomising stuff comes from.

        Would be interesting to see how the detail compares to the original Bayer design.

  • Berneck

    The more I read about this camera, the more it sounds like what I’ve been waiting for. Please let it be fast! I guess we’ll have a better idea by Monday.

    I wish Oly/Pana would do something like this! The m43 format really is the perfect balance of size and image quality…

    • brian

      I have a feeling it’ll be less than fast. This post already states that the tech to avoid moire comes at the expense of time.

      I’m expecting the speed to be the disappointing factor with this camera. Along with poor software/menu structure which I believe Fuji already has a poor reputation for.

      • brian

        In fact, wasn’t the lack of speed one of the main drawbacks to the s3/s5 dslr’s? And here they’re already stating that the new sensors colour reproduction comes at the expense of CPU time.

        • Just think for a moment… if they’re saying it’s going to be slow it’s going to be given as an option. So I’m sure you can turn it off when you’re machine-gunning for your life.

          And seriously… anyone who can’t learn camera menus should stick to a hobby like gardening.

  • I’m trying to decide which camera to sell first so I can purchase the X-Pro1.

    I got–all of them!

  • Mark

    Fuji’s solution for moiré is highly interesting.

    What Fuji factually did by “randomizing” the pixel pattern, is using a varying sample frequency for different spaces.
    Moiré is created because the resolution of detail in the real world is too high for the sensor (it exceeds the so called Nyquist frequency). The camera then think it sees a different frequency which does lie in the frequency range of the camera (aliasing) and it displays this frequency. However, this frequency is totally faulty and we see moiré .

    Now, with Fuji’s solution, they don’t overcome this problem, they just minimized the effects of aliasing. Because the sample frequency is variable, so is the Nyquist frequency. This, in turn, when aliasing occurs, randomizes the faulty frequencies the camera sees. Because the faulty frequencies are randomized, there won’t be a pattern in the aliasing objects, and we won’t see moiré.

    At least, this is how I think it will work =P.

    Though the AA filter can now be removed, it brings more problems with it as well. In order to convert a picture from RAW to JPG, you need a constant sample frequency. So, the camera will have to interpolate a constant frequency using the given randomized frequencies. This leads to more CPU time as already noted by PR guy.

    Another potential drawback: don’t expect conventional RAW files from this camera. All current RAW converters only work with constant sample frequencies. Either the camera won’t support RAW, or a new format/ compatibility will have to be created for this camera.

    • Yamuk

      Interesting read 🙂

    • LGO

      My concern is whether a user would be limited to using Fuji’s software or whether he can use ACR/PS and Lightroom.

      • Mark

        I do suppose so, as Adobe Camera Raw also supports sensors with a non-classic architecture now (like Foveon)

        • Foveon files don’t need any interpolation.

          This is not the first time a non-Bayer RGBG, non-Foveon (non-layered) design has come about. E.g. Fuji Super CCD, Sony RGBE, Kodak CYYM, etc.

          As far as I know, ACR and DNG have supported all of them if the camera produced raw files.

    • A few bad speculations up there…

      1. The said “random” arrangement only mixes up the colour filters within a 6×6 neighbourhood of photosites. So it’s not quite a variable sample frequency where you’d get more resolution from one spot and less from another.

      2. If there’s a hit on the speed of raw conversion, it’s because the converter will need to loop up (in a table) which colour each photosite is. Or the colour filter information (i.e. whether R, G, or B) should be attached to the values of each photosite in the raw file. This is going to require another 2 bits of information per photosite. So, approximately 4MB extra per raw file.

      An easier way to handle this is at the time the camera writes the raw file out. The camera can have the look-up table and write out the values to the raw file in a de-randomised sequence like RRGGBB. This will make it a lot easier to write raw converters for this format.

      But even standard raw converters also involve a lot of optmisations to increase detail and reduce artifacts. Fuji’s design might increase the minimum computations needed to convert a raw file but it might not need the extra processing necessary for Bayer artifacts, moire, etc.

      3. Raw processing software get updated for different cameras you know?

  • Lauro

    Be afraid of the price

  • I’m a big fan of Fuji but I’m very skeptical about the effective resolution capabilities of this design when compared to a Bayer sensor of the same rated MP resolution.

  • Back to top