Noise pattern for the main camera dng/jpg

  1. teo_paul
    Donut Jun 18, 2019

    teo_paul , Jun 18, 2019 :
    Hi guys,

    I am experiencing some strange noise pattern on the main camera photos, mainly visible when you look at the DNG files since the JPGs offer less details. It looks like a "maze" connecting dots which otherwise should look like individual noise dots.
    Are you seeing the same thing I am seeing? because sometimes this kind of pattern can only randomly affect some sensors (see attach - a 300% zoom on a jpg exported from a Lightroom processed dng).
    noise test.png The zoom is done only to make the noise more visible for everybody, not that we should inspect the photos at 300%.


    Attached Files:

  2. stark2991_PL
    KitKat Jun 18, 2019

    stark2991_PL , Jun 18, 2019 :
    And? It's probably because of the pixel size and algorithms that join pixels together.

  3. teo_paul
    Donut Jun 18, 2019

    teo_paul , Jun 18, 2019 :
    That's for sure, I just wanted to see if someone else noticed that. Also this maze pattern might end up in "painted" details in jpg files.
    Most of the users complaining about lack of details and sharpness in the photos do not realize that those things are the sum of small issues adding up.
    Raising awareness is always a positive thing.

  4. stark2991_PL
    KitKat Jun 18, 2019

    stark2991_PL , Jun 18, 2019 :
    Most of the users don't know what noise is and why it is present on a photo.
    I think that Photography section is better for such discussions.

    otto2 likes this.
  5. hotsauce
    Lollipop Jun 18, 2019

    hotsauce , Jun 18, 2019 :
    People would be able to better assess the situation if you include the full size picture and the camera settings.

    For example this is a picture taken in very dark restaurant, it is with the main camera in normal mode at 1/10s and iso 1600. As you can see there are no checkered pattern.


  6. teo_paul
    Donut Jun 18, 2019

    teo_paul , Jun 18, 2019 :
    Noise is not all the time visible, especially on the standard jpg where usually the camera algorithm removes it during the processing phase.
    But for the sake of it, here it is the full photo (I see the forum page compresses it heavily) .

  7. pgrey
    Ice Cream Sandwich Jun 18, 2019

    pgrey , Jun 18, 2019 :
    The thing is, I bet most users (a VERY high %) would think that photo is "just fine", and never see the noise anyway, because it's not really what the subject of the photo is focusing on.
    I think it's great, to keep improving "noise reduction", to a point, but at some threshold you get crazy-sharp edges, in the place of the dithering that the algorithms are doing now.
    This is where the sensor-processing gets down to somewhat of an algorithm/art interpretation, if you ask me, and there are very few people in the world, in the greater body of s/w and h/w engineers, that can refine these.
    I'm sure there are people working on all sorts of "AI sample algorithms", to try and mitigate this issue, but this is going to take time, or hiring the "right engineer or two", IMO.

    Personally, I say use a camera, that has a great sensor, and a great algorithm, that's well-proven, if you want to take a great photo. I use my mobile phone as a "it's all I brought" kind of camera...
    I've said it before, between the sensors and the optics-depth, you're just not going to make a "real digital camera" out of a cell phone, with current tech. You can get a really good "interpretation", but not a consistently great one, IMHO.
    Even a 100,200, whatever, sensor, isn't going to solve this, it's just going to "reduce the likelihood", IF the algorithm is good.

  8. pgrey
    Ice Cream Sandwich Jun 18, 2019

    pgrey , Jun 18, 2019 :
    @hotsauce Yep, that's another good example. Zoom on the napkin, if you really want to see a "big mess" of that.

    There's pretty much always a trade-off with this kind of stuff. You can have fast/good/cheap processing of noise, choose 2, or maybe 1, for mid-level phone devices.
    There's a reason that cameras have really high-quality processing chips, and optics, and why they will excel here.

    Low-light is trickier, still, if you say "low light processing" in the above, you might only get to choose 1, or maybe 1.5, depending, in a mobile device, even most flagships.

  9. teo_paul
    Donut Jun 20, 2019

    teo_paul , Jun 20, 2019 :
    A new, relevant and important review has just arrived for Oneplus7 Pro.
    Andrei Frumusanu from Anandtech has always deep analysis about the high end phones and I was waiting for his opinion on this phone.
    As I expected, the screen and performance are top notch for mid 2019 and there is no competitor for the moment in these areas. But also, as expected, the camera performance is not up to par with the top competitors (not even with the competition that uses the same sensor).
    It is actually the first review that induces the idea that Oneplus is actually echoing the photo representation of the Google Pixel, which is not necessarily a good thing. The Google Pixel won the appreciations for other things like night mode, hdr and detail accuracy, not for the colors, brightness or some other aspects of the photo. This direction seems to be a pretty wrong one since the last survey that OnePlus conducted related to photo preferences revealed that users actually like the Samsung S10+ look instead of the others, which means for me bright pictures, punchy/saturated colour, bright shadows and overall a HDR approach that shows a better dynamic range coverage.
    I am curious if Oneplus will steer the photo algorithm towards the new direction chosen by the users (btw, why wasn't this test conducted prior to OP7 appearance?).