Greeting my fellow Kine-nerds and welcome to the first of 2019’s tips and tricks blogs. To ease us back into the new year we’re once again going to tackle clarifications. We’ve covered most of the other clarifications in blog #36, #37, #38, #39, #40, #41 and #42 (there are a lot of them!). Today we’ll be dealing with stabilize and temporal median.


When you get video from bodyworn cameras or phones you often want to stablize them. This function allows you to correct the camera shake, rotation and changes in angle and zoom level. It woks similarly to Perspective Crop but allows you to fine tune the selection for each frame. You should select some clear features in the video and click on them to assign stabilzation points, four points in total. These points should, if possible, be visible throughout the video segment to be stabilized. The stabilization points may be fine tuned for a number of frames as required.

Temporal Median 

Temporal Median combines many video frames to filter out high gain noise. This filter, in combination with Histogram Edit, can be useful for recovering license plates or logos in dark and noisy videos. It works by statically modelling the scene and finding the median value of each pixel over time. This process naturally cleans up non-moving objects and filters out any changing pixels or moving objects. This works best when the video dark, but there is some light present. Completely dark videos cannot be clarified using this method.

To recover a logo or license plate in this way, go to the forensics tab and find a section of video in which the object to be clarified is non-moving. In the clarification tab, select the clip and add the Temporal Median clarification. The default settings are Median and a Combine Frame number of 30. This is usually optimal.

The default settings are Median and a Combine Frame number of 30. This is usually optimal. The mode options refer to the pixels. So the max value will return the brightest pixels from all the frames. The combine frames sets the number of frames taken.