Why the Amped FIVE Filter Sequence Matters in Forensic Video Analysis

Table of Contents

Reading time: 5 min

If you have attended an Amped FIVE training (if you haven’t yet, here is the list of upcoming classes), then you know how much stress we put on the filter sequence in Amped FIVE. The footage you’re working with is the product of a long acquisition and processing chain, which usually introduces several artifacts. The most reasonable way to go is to compensate for these artifacts in the reverse order. If you like math, you can find some reasons supporting this choice, which is also summarized by the drawing below. As you can see, going from X to Z involves passing first through f and then through g. Going back from Z to X requires going through the inverse of g first, and then the inverse of f.

Mathematical diagram illustrating function composition and their inverses, showing sets X, Y, and Z connected by functions f and g, with forward mappings on the top row (f: X → Y, g: Y → Z) and inverse mappings on the bottom row (f⁻¹: Y → X, g⁻¹: Z → Y)
By Jim.belk at English Wikipedia – Transferred from en.wikipedia to Commons., Public Domain

If you’re more on the practical side, it’s still quite intuitive: if you first put on your socks and then your shoes, you’ll have to take off the shoes first and then the socks. In this trivial example, “take off” is the inverse function of “put on”. You need to take off clothes in the reverse order of how you put them on! This logic directly applies to the Amped FIVE filter sequence. Using filters out of order can compromise the integrity of the enhancement process.

We should thus keep in mind that, in general, the acquisition chain introduces a series of defects. These typically appear in the following order:

  • Scene-related issues – such as bad perspective, distant subjects, or obstructed views
  • Camera lens limitations – including distortion, chromatic aberration, and other optical flaws
  • Camera sensor defects – like image noise, poor contrast, or low dynamic range
  • Encoding artifacts – such as interlacing, compression blocks, or pixelation
  • Acquisition process artifacts – for example, faulty proprietary players exporting video with the wrong aspect ratio or reduced quality
  • Defects due to the analyst’s viewing system

Visual diagram illustrating the digital video evidence workflow, from a crime scene captured by a surveillance camera, to storage in a DVR system, acquisition via forensic hardware, and final viewing on a computer monitor for analysis

Okay, okay., but will it make any difference if I get it wrong? If you’re one of those who needs to see things in practice before believing them, let’s take a look at the difference it makes to use the proper filter sequence.

We’ll make three examples with very simple chains:

  1. deblurring and perspective
  2. deinterlacing and stabilization
  3. crop and undistort

Deblurring and Perspective

Let’s better understand the importance of the Amped FIVE filter sequence. We’ll walk you through some real-world scenarios that show how much the order impacts the outcome.

Let’s work with this boarding pass: we need to take out the passenger’s name and flight number.

Blurry close-up photo of a printed Ryanair boarding pass on a desk, showing flight details and barcode partially visible

We see the image is strongly blurred, and its perspective is not favorable. “Easy to fix, with Amped FIVE!”, I can hear you say, and you’re right! We just need the Optical Deblurring filter (under the Deblurring category) and Correct Perspective filter (under the Edit category). But in which order should we use them? Let’s think about it: in the acquisition chain, the position of the document with respect to the camera is at the very beginning of the story, it’s just the way the things were in space at that moment the picture was taken. The blurriness came after: when light traveled from the paper through the lens, the lens failed to properly focus it on the camera sensor, turning something which should correspond to a single pixel into a circle in the image.

Diagram illustrating how light rays from a point on an object pass through a lens and form a circle on a sensor, explaining the concept of optical blur in imaging systems
An example showing why poor focusing turns points to circles in your image.

Therefore, we will first need to compensate for the last step (optical blur) and then correct the perspective. The image below shows the result if you use the right order (top figure), and the result if you use the wrong order (bottom figure). Indeed, if you first change the perspective, each pixel is no longer a circle, it becomes an ellipse, and Optical Deblurring will hardly restore it.

Distorted and enhanced comparison of a Ryanair boarding pass showing passenger details, seat number 17B, and destination London Stansted

Deinterlacing and Stabilization

Let’s now turn to this other case. This time we need the license plate, but we see that the video has poor resolution and is interlaced.

Rear view of a red Fiat Panda driving on a gravel surface captured by surveillance camera, showing horizontal interlacing artifacts

We can use the Deinterlace filter to remove the “jagged lines” effect, and then use Local Stabilization followed by Frame Averaging to get the best possible view of the license plate. Again, which order should we choose? The poor resolution is due to the distance of the car from the camera, while interlacing has been added well later in the acquisition chain. Therefore, we first need to apply Deinterlacing and then stabilize the video and average frames. Like we did before, we show what happens if you use filters in the right order (top image) or if you reverse and put deinterlace after stabilization and averaging (bottom image): quite a difference, isn’t it? This example reinforces how crucial it is to follow the correct Amped FIVE filter sequence to avoid visual degradation.

Comparison of two enhanced license plate images showing the rear plate DT210MM on an orange Fiat vehicle, demonstrating improved readability through forensic video analysis techniques

Undistort and Crop

The last example involves taking out a good image of the glass skyscraper on the right side of the image below: it doesn’t take much expertise to notice there’s a strong lens distortion in this picture.

Street-level view of Trump Tower in Midtown Manhattan, New York City, captured with a fisheye lens, featuring surrounding skyscrapers, city buses, yellow taxis, and pedestrians on a busy urban street
Credit for original image: https://pxhere.com/en/photo/596533

Luckily, this is no problem with Amped FIVE! We just need the Undistort and the Crop filters to reduce the distortion and keep only the object of interest. Which order should we opt for? This time, it comes handy to know the underlying assumptions of the Undistort filter (which, of course, are explained during training and documented in the Filter Reference Guide, available under the Help menu). It reads:

The center of the image must coincide with the optical axis of the lens (i.e. the filter cannot be applied to cropped images).

Therefore, we must first remove distortion and then crop our object. We are again showing below the result of applying filters in the right order (top image) and in the wrong order (bottom image). Choosing the right filter sequence in Amped FIVE here ensures the filter assumptions are met and the final image remains geometrically correct.

Once more, a huge difference!

Comparison of architectural distortion correction on a tall glass skyscraper in New York City; the top image shows the building with minor distortion, while the bottom image illustrates enhanced fisheye lens distortion with exaggerated curvature of vertical lines

Conclusion

As we’ve seen in the examples above, getting the filter sequence in Amped FIVE right can mean the difference between clear, actionable footage and a missed opportunity. So next time you’re processing video evidence, take a moment to consider the logic behind the sequence, it’s worth it.

Table of Contents

Share on

Subscribe to our Blog

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Subscribe to our Blog

Related posts

Amped Engine Updates

Amped DVRConv and Engine Update 36977

Reading time: 2 minThis latest release introduces support for new proprietary CCTV formats and extended support for existing formats and variants. We have also added a new audio

Read More »