Posted by: recordingsofnature | March 31, 2012

Antialiasing filter for DSLR video

This post does not deal with audio, but with video as a change. I have some ideas of combining audio with video in some  future posts.

This is a demonstration of a DIY method to effectively reduce moire and aliasing problems occurring with  common DSLR cameras. The method uses a simple uncut ophthalmic lens.

Moire patterns can be quite visible with DSLR cameras for certain scenes. The chosen scenes in the video; the ISO resolution chart, wooden table and the brick wall, are some of the very severe moiré-cases I have encountered with the Nikon d5100 camera. Watch out for flickering interference patterns on the table and brick walls!!!

[Update 2012.09.25, New post with more detailed sharpness measurements here].

[Update 2012.12.04 About filter fitted into a standard screw frame]

[Update 2016 April, DIY variable anti-Moré filter method for DSLR]

While the popular DSLR cameras offer great lens and sensor qualities, they are normally suffering from specific drawbacks. One of these are the aliasing, Moiré interference patterns. From time to time it really disturbs the video quality, by introducing high frequency noise and artifacts. This is especially visible with textiles, brick walls, wires and other regular structures. The aliasing gives noise in the image, and stresses the video compression.

This wikipedia shows good illustrations (especially the four blue fractal pictures) of how the image quality can be improved with improved anti-aliasing. When the aliasing is present the image appears grainy and of lower resolution. With super anti-aliasing the same image gets an incredible smoothness, depth and detail as if the resolution was much higher.

The nature of moiré and aliasing artifacts are due to line skipping at the image sensor, and it actually only causes problems in only one direction. Each line is sampled fine, but when jumping to the next line the area between the lines are not used on the sensor. Horizontal lines get over-sharp. A good solution is to blur the image before it hits the sensor.

The idea here is to use an uncut eyeglass lens with a little bit of astigmatism (cylinder lens) only. This modifies the existing focus point with two new ones at two different focus depths, each blurred in either vertical or horizontal direction. By putting the lens in front of the camera lens, with the cylinder axis at an angle of 45 degrees to the sensor in the camera, there will always be a little unsharp when you try to focus, and this is useful to remove the moiré patterns.

anti moiré filter uncut eyeglass lens

Camera and eyeglass lens

The video shows a footage with and without the cylinder lens. Used here is a +0.5 cyl., which seems to be the optimal solution for the lens setup, 18mm + aperture 3.5 on the Nikon d5100. As seen, the moire is almost gone, without so much other degradation of the image quality. Edge sharpness is ok. [Detailed sharpness measurements are now available here.]

It is not the most user-friendly solution though, since the blurring effect must be adjusted to an exact amount – not too much and not too little. This effect depends on the aperture and the focal length of the lens. By increasing the depth of focus the blurring will be less. And by zooming in the blurring with increase with the zoom factor. So it is less suited for zoom lenses.

In the video I use autofocus to be sure to have the sharpest image (for which the moiré is strongest), though normally i would use manual focus with this camera. I have uploaded the video unedited and straight from the camera to preserve the original quality as much as possible. Only youtube audio swap.

(update April 3th 2012 )——————————————————–

Some April spring video from Vest Amager, Copenhagen, Denmark – testing the add-on anti-aliasing lens:

The file is uploaded raw without any editing.
– and hmm, the auto converted 3D version is quite interesting.

If any one interested, this is the complete list of video settings:

  • Kit lens 18-55mm, @ 18 mm
  • ND 16 filter
  • Anti-aliasing filter 45 degrees, 0.5 cyl (see previous)
  • Manual focus
  • VR on, tripod
  • 30 fps HQ 1080p
  • A-priority, aperture 3.5
  • Fixed exposure, 1/50 sec, 3.5 A, iso 125, equal ~0EV compensation
  • WB: cloudy +2 warmth (ND filter compensation)
  • Neutral picture profile: sharpness 3, contrast -2, brightness +1, saturation +1
Advertisements

Responses

  1. how can i get this filter? Do you sell one?

    Like

  2. my address below!

    Like

  3. […] article follows up on the antialiasing filter method described in the earlier post  Antialiasing filter for DSLR video, for reducing or eliminating moiré video artifacts typical for DSLR HD videoing, -and particularly […]

    Like

  4. […] (size 49, 52, 55, 58, 62 and 67 mm). See more info about the filters in earlier posts, here and here.  A small-scale piecewise assembly has been established. If interested, please contact me directly […]

    Like

  5. Hello, thank you very much for the ingenious find. I think this filter would be better suited for cokin type slide-in filter system, because screwing-in filters is a bit cumbersome. I have a question, why the 45 degree angle? We only want vertical blurring because line-skipping occurs only in the vertical direction. Horizontal direction should be okay. So isn’t 90 or 180 better suited for the application?

    Like

    • Thanks. It is true, that a slide-in type of filter could be more flexible to use. The main issue is that in order for the filter to work well, it has to be as close to the front element of the camera lens as possible. A distance of few centimeters will unfortunately produce a highly distorted image.
      Regarding the 45 degree angle: The filter works by turning the original focus spot into two perpendicular lines at two new focus depths (astigmatism). In between, the focus spot is now a larger elliptic/circular spot. So when the angle is 90 degrees, (which is actually the same as 0, 180 and 270 degrees) you can surely get a vertical focus point, but there will also be a focus distance where the focus spot line is horizontal, which will still allow aliasing in the image. Finding the right focus distance could be as difficult as manual defocus (But in some cases with autofocus it might also work?)
      When the angle is 45 degrees, there will never be a perfect focus. There will always be both a vertical and horizontal blur component. When focusing, the sharpest focus will be when the focus spot approaches the circular spot, which also gives an optimal antialiasing effect.

      Like

      • Thank you for the clear explanation. I believe the circular slot of cokin filters is pretty close to the lens. Cokin A series filters are 67mm. Eyeglasses lenses are 65mm. So it could be a close fit. But the curvature of the eyeglass lens could be a problem, I don’t know. I am a bit busy right now but I will try it and let you know about the results when I have time.

        Like

  6. […] see more at https://recordingsofnature.wordpress.com/2012/03/31/antialiasing-filter-for-dslr-video/ […]

    Like


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Categories

%d bloggers like this: