Tips on Time-lapse Flicker
Flickering is a common problem in time-lapse or stop-motion footage captured using a
digital still camera. Frame by frame variations in lighting or in exposure lead to perceptible
brightness fluctuations, or flicker. Flicker is more of a concern when using GBTimelapse for
videography than for monitoring.
Some amount of flicker is almost unavoidable. Thankfully it can usually be removed from the
image sequence with a tool like GBDeflicker. The algorithm behind GBDeflicker is powerful and
technically complex. The software creates a luminance histogram for each video frame, and then
adjusts the image histogram, frame by frame, to minimize flicker. GBDeflicker can usually
eliminate any flicker as long as there is no significant clipping.
Sources of Flicker
There six primary sources of flicker.
| Flicker Source | Remedy |
| Aperture flicker | Use a smaller f-number |
| Shutter flicker | Use a slower shutter speed |
| Av stepping | Use a constant Av value |
| Tv stepping | Use bulb exposure if possible |
| Natural flicker | Use a GBDeflicker sub-rectangle |
| Lighting flicker | Avoid fluorescent lighting or use long exposure |
Cause #1 - Aperture flicker
Remedy - Use a smaller f-number
Because the camera iris is a mechanical device, there are differences in the exact size of
the aperture for successive actuations. These small differences in aperture result in
frame-to-frame luminance variations that show up as flicker when played back. Aperture
flicker seems to be less evident with smaller f-numbers (larger apertures).
Split image of two successive frames
The left and right halves of the image to the right are from two successive
frames captured three seconds apart with the same camera settings. The right
half is noticeably brighter than the left half. I loaded both images into
Photoshop and compared the histograms. They looked almost identical, but had
mean values that differed by 4.5 percent (77.50 versus 81.01). Although the
clouds did move a tiny bit between these images, it is not enough to account
for that large of a difference. Nothing passed in front of the sun, and I
assume the sun's light output did not vary between the shots.
How can a digital camera produce brightness variations in successive images of
the same subject with identical camera settings? You would expect a digital
device to give identical results from identical settings; but that is not the
case.
A DSLR camera is not electronic. It's iris is mechanical and mechanical devices have intrinsically limited
precision and repeatability. For each shutter press, the camera activates a
mechanical device that moves the iris to the desired aperture. The aperture
cannot be exactly the same size every time because of friction and other
factors so there are differences in exposure. You would expect a higher quality
camera to have a more precise aperture mechanism.
To further investigate, we shot twenty successive frames of the same scene at
different aperture settings and plotted a graph of the average luminance. Each
colored curve in the graph below is for a different setting.
Cause #2 - Shutter flicker
Remedy - Use a slower shutter speed
DSLR cameras also have a mechanical shutter, so there will be differences in the
exact duration of the shutter for successive actuations. These small differences
in duration result in frame-to-frame luminance variations and produce flicker.
Using a slower shutter speed is better because a small timing error is less
apparent with longer shutter times
Cause #3 - Av stepping
Remedy - Use a constant Av value
Typical camera Av settings are spaced at 1/3 stop intervals (f4, f4.5, f5.0, f5.6).
As the Av value is changed, there are small steps in the luminance because of the changes
in the amount of light passing through the iris.
Cause #4 - Tv stepping
Remedy - Use bulb exposure if possible
Typical camera Tv settings are spaced at 1/3 stop intervals (1/10, 1/13, 1/15, 1/20 etc.).
As the Tv value is changed, there are small steps in the luminance because of the
changes in the amount of light passing through the shutter. Because bulb exposure times
can be set with a precision of 1/1000 of a second (2.100, 2.101, 2.102, etc.), smaller,
more gradual changes produce less flicker.
Above is a screen shot from GBDeflicker showing the luminance graph during a
daylight-to-night AutoRamp timelapse. The first part of the graph shows a “saw-tooth”
luminance pattern as the Tv value was stepped down in 1/3 f-stop increments. After
switching to Bulb mode, the luminance curve is smooth.
Cause #5 - Natural flicker
A timelapse video of passing clouds may include shadows being cast on the ground.
Although the sky may have a slowly changing luminance, the luminance of the ground may
fluctuate rapidly because of the shadows.
Remedy: Use GBDeflicker sub-rectangle If there is unnatural flicker in the sky it can be
removed by using GBDeflicker to analyze only a sub-rectangle of the sky in each image.
Remedy: Use AutoRamp instead of Program mode
Because a camera’s Program mode determines the exposure at the moment the shutter is actuated,
variations caused by cloud shadows will affect each exposure. When using GBTimelapse AutoRamp,
the exposure is determined from a moving average over a number of past images – this reduces
the influence of cloud shadows between successive frames.
Cause #6 - Lighting flicker
Any artificial lighting that is connected to the power grid can produce flicker.
The 60Hz (or 50Hz) power frequency produces a corresponding fluctuation in the light output of the lamps.
Fluorescent lamps are the worst. From Wikipedia, "Fluorescent lamps which operate directly from mains
frequency AC will flicker at twice the mains frequency, since the power being delivered to the lamp drops to zero twice per cycle.
This means the light flickers at 120 times per second (Hz) in countries which use 60-cycle-per-second (60 Hz) AC,
and 100 times per second in those which use 50 Hz. "
Incandescent lamps also flicker. The light produced by the lamp depends on the current
flowing through the filament. Since the current is varying with a sine wave, the light
output will vary with a sine wave too. Lamps with larger filaments have more thermal mass
and produce less variation. In general, lamps in Europe show more flicker because
they use thinner filaments to match the higher grid voltage.
Remedy - Use long exposure An exposure time much longer than the power frequency will help average out the light fluctuations.
Remedy - Avoid fluorescent lighting The most obvious is fluorescent lighting with flickering that is apparent to the naked eye.
Minimizing Deflicker Problems
GBDeflicker is a tool for removing time-lapse flicker. It is available as an Adobe Plugin or as a standalone application.
GBDeflicker removes flicker by adjust the luminance of each frame in an image sequence to minimize the frame by frame variations.
It does this by mathematically shifting the histogram of a frame left or right (darker or brighter).
GBDeflicker works very well as long as there isn’t any significant clipping of the image data. Clipping occurs when some
image pixels are at the maximum possible value. This part of the image is likely overexposed resulting in loss of image information.
When pixels values are clipped, the histogram cannot be shifted correctly because there is no way to calculate the corrected value of an overexposed pixel.
Looking at the graph above, it’s not possible to know what the blue channel should look like if it were shifted to the left.
The clipped information has been lost.
Avoid Clipping
Use Program Mode with negative exposure compensation – A negative exposure compensation
value is less likely to produce overexposure and clipping.
User a lower target luminance in AutoRamp mode. A lower target luminance produces is less likely to produce overexposure and clipping.
Use Neutral Picture Style
Canon cameras have a Picture Style setting that determines how the raw image data
is converted to a JPEG image. The Standard setting increases the saturation and may
cause unnecessary color channel clipping (particularly the blue channel).
The Neutral setting seems to be the best setting to avoid color channel clipping.
Two images below show the same scene shot with Standard and Neutral picture style.
You can see from the histograms that with identical exposures, the Standard saturated
the blue channel to the point of clipping.
You can see in the above histogram how the “Standard” picture style boosts the value of the blue channel pixels.
This is a “nonlinear” process that is difficult to deflicker.
Above you can see how (at the same exposure) the Neutral picture style has not clipped the blue channel. Using the Neutral picture style is
very important when shooting timelapse of blue skies. It is less important with more balanced images such as grey clouds.