The Science Behind Number Plate Enhancement

With CCTVs at every street corner, vehicle number plates are an important part of every investigation. However the low-cost CCTV devices employed tend to produce somewhat poor quality results.

Additional factors such as lighting, distance, angle of capture, camera resolution etc all contribute to end results which leave the investigator guessing as to the exact number on the licence plate.?

While there are a lot of technologies out there that help in enhancing number plates, it is important to understand that some cases are a lost cause and the plates cannot be enhanced no matter what. Rather than expending effort, time and resources on enhancing plates that will produce no result, we need to be able to quickly identify/differentiate those plates that can be enhanced and those that cannot.

The first step to this is by understanding the underlying science.?

Let’s start by reiterating that all pictures and by extension CCTV videos of number plates are composed of pixels. A number of pixels placed one on top of the other in a particular fashion form a number. To understand this better, we need to start with a pixel grid. We can start with a 20 x 20 grid and take it from there.

We begin by creating a readable number in this grid. We will do this by stacking pixels (we will consider one box of the grid as a pixel) one on top of the other in a fashion to make it look like a number. In figure 1, we can see that we can create a number with as little as a 5 pixel grid.?

Figure 1 - Five Pixel high grid depicting the number 8

However, while this does look like a digit, it looks nothing like a number from a number plate. To do a numberplate justice, we need to stack and properly arrange pixels in a 12 pixel high grid to reasonably approximate a number plate quality number.

Figure 2 - A bigger grid now makes it possible to see a better approximation of the digit 8

So, what does this mean? As we can see it takes a minimum of about 12 pixels stacked one upon the other to build a proper readable number. Hence if the image or video frame has a resolution which is less than 12 pixels in height, there is a very low/no? possibility of forensically arriving at the license plate number no matter the extent of video enhancement.?

Hence a quick check to see if the captured CCTV video/image is at least 12 pixel tall will help save hours of enhancement efforts and free up the analysts time to focus on other videos that are more likely to produce results.

I call this “The Rule of 12“.?

Next time, please use this “Rule of 12” to determine if it is worth investing your time in enhancing video footage to get at a number plate.

Good luck and Happy Enhancing!