I recently had my laptop stolen, with lots of pictures of my family, research work and other very important stuff. Of course the theft has been reported, but I think people here are more competent in image processing compared to the local police station :)

Video Recordings

The stealing was captured by a security camera, original .dav files can be downloaded from here:

Converted .avi files can be downloaded (or played in browser) from here:


In the file named ch10 (full name: HCVR_ch10_main_20160501143700_20160501143947) the thief's car enters the frame from the right at 0:48 . In the file named ch13 (full name: HCVR_ch13_main_20160501143811_20160501144205), in the first 2 minutes, the thief's car enters the frame from the right, makes U-turn, steals my stuff, makes a U-turn again and go.

What has been tried?

I tried cropping a few frames around the license plate, stretch so that the plate will be on same position and size on every frame, and averaging. The result wasn't better from the original frames :(

How could I get the number?

  • $\begingroup$ Welcome to DSP! Good luck with recovering your stuff, especially the pics. :-( $\endgroup$
    – Peter K.
    May 6, 2016 at 17:57
  • $\begingroup$ Can you post a few pics of the crops? I had my car stolen once along with a boot full of climbing and sporting equipment. It really sucks. $\endgroup$ May 15, 2016 at 0:17
  • $\begingroup$ Your approach is faulted, I think. Image processing might not be the best option, since the streams are impaired by artifacts, motion, and low resolution. Not to speak of the possibility that you yourself acted illegally by making those videos publicly available (yes, in some countries this might be a problem...). Better: The guy obviously is on the phone. You have an accurate time when and where it happened. Let the police correlate the active callers in the cell with the owners of the brand/type of car. Probably, this is more fruitful than chasing wobbly pixels. Future: Encrypt and backup! $\endgroup$
    – M529
    May 15, 2016 at 9:40
  • $\begingroup$ Did you finally recover your stolen good? $\endgroup$ Dec 18, 2018 at 22:12
  • 1
    $\begingroup$ @LaurentDuval :( no $\endgroup$
    – User
    Jan 3, 2019 at 9:01

2 Answers 2


Averaging may not be the best option, due to motion compensation in video frames: one number (say $8$) may be extrapolated for another (say $3$) on a different JPEG block. It might be interesting to dig into the compressed video format, check motion vectors, but this requires a little bit of hacking.

Some software which can help:

Unfortunately, while your videos are pretty good, the problem sounds quite difficult. For alight side, I'd like to share the following picture, that tells about issues in image forensics:

image processing CIS

  • $\begingroup$ I had a dig at the videos frame-by-frame and I could barely eyeball WHERE the plate was, let alone make out characters. I might need to use the non-AVI format. $\endgroup$
    – Peter K.
    May 7, 2016 at 16:14
  • $\begingroup$ What would moderators do! $\endgroup$ May 7, 2016 at 17:15
  • 2
    $\begingroup$ Phone a friend? $\endgroup$
    – Peter K.
    May 7, 2016 at 21:56

There is a company named Signalscape who has a line of products called StarWitness (http://www.starwitnessinfo.com/) that do forensic video analysis that are available to law enforcement agencies (only). I will try to get a hold of the director of the StarWitness group to determine if there is some course of action you can take that may assist you in recovering your stolen items. Perhaps you can ask if your local law enforcement agency has access to StarWitness...you never know. StarWitness Video has a lot of tools that can enhance video, including image stabilization. Unfortunately, Laurent's comic brings to light the major issue - you are at the mercy of the resolution and any compression.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.