When considering forensic evidence and its applicability in the court system, there can potentially be a few issues. For example:
-
Evidence can be big or immovable. It may not be practical or even possible to attempt to analyze an entire vehicle or the ground a car drove across while working within the confines of a crime laboratory.
-
Forensic scientific practices must be reproducible by opposing experts. If work is performed directly on an evidence item, it makes it difficult for another expert to complete their own analyses on the same evidence.
-
Some evidence is transitive or volatile, so it’s important to analyze it while it is accessible.
The easiest solution to mitigate each of those concerns is to document the evidence photographically and work from these photographs. The focus of this article is not to teach how to take these images. Instead, the goal is to illustrate the necessity of understanding what makes up an image and the importance of using a proper image processing utility after capturing the images.
What makes an image?
An image is a digital representation of the light energy captured by the sensor of a camera. The same basic principles apply, regardless of whether the camera captures video or still images. Capturing those points digitizes the analog world. Images are divided into pixels (picture elements), which make up the resolution (rows and columns) of a digital image. The more pixels available, the higher the resolution of the image. Because rounded shapes cannot be fully represented with square pixels, the camera’s internal processing is forced to make some decisions about how to handle edges and details within the imagery. This processing decision is referred to as quantization.
In most cases, the influence of the mathematical calculations performed within the camera has an indirect relationship with the image resolution, therefore it is preferable to have as high of an image resolution as possible. Most comparison-quality images are taken with this concept in mind, even if the term is unfamiliar. Photographers are taught to fill the frame with the subject matter and to use a RAW or otherwise uncompressed image format. In filling the frame, as many available pixels of the resolution as possible are utilized for the important image details. This prevents the relevant minutiae from being relegated to individual pixels for accurate representation. Additionally, by using an uncompressed image format, less damage is applied to that high-frequency detail due to quantization.
Why is image clarification necessary?
While it is preferable to begin with a properly captured image, many factors might need to be accounted for after the images have been recorded. Some things might have been controllable during the image capture, but others might not. Certain lighting conditions, blur, distortion, pattern interference, and other factors can be adjusted or corrected for after the fact, and sometimes may be the only route to usable evidence.
The first consideration for detail clarification is to have an optimal amount of contrast between the evidence and the background. Whether these details are fingerprint ridges, shoe impressions, or even license plate characters, the separation of an area of interest from the background area surrounding it is important to fully assess minute characteristics. Contrast can be emphasized by equalizing a histogram, adjusting levels, or applying a Homomorphic Filter (Image 2).
Image 2. © 2025 Amped Software
Disruptive background patterns can be reduced by viewing only specific color channels or by analyzing a fast Fourier transform and eliminating extraneous frequency information (Image 3).
Image 3. © 2025 Amped Software
It might be beneficial to analyze the image in grayscale (Image 1) or, in some cases, as a negative so that the dark and light characteristics of the imagery are inverted (Image 4).
Image 4. © 2025 Amped Software
It is also essential that items requiring physical matching (e.g. ridge detail, shoe impressions, etc.) can utilize a scale within an image to calibrate it 1:1. This means one inch in the image equates to one inch in real life. This calibration can only be properly done if the plane of the item captured in the image is parallel to the camera’s sensor. If necessary, geometrical calculations can be performed to correct lens distortions due to images taken at an angle.
This work should be performed in a non-destructive, layer-based image processing approach to protect the integrity of the original image throughout the process.
As questioned items are analyzed, annotating the image with features of interest, identifying marks, or any other necessary documentation might be necessary. Many individuals in practice print imagery and perform these markups on the paper representation of the images. However, if an image processing tool allows for easy markup in a separate layer, this can ease the workflow.
After analyzing unknown items, displaying the unknown side-by-side with a known one allows for easier comparison between the two. In some instances, it might also be beneficial to view certain images as an overlay with another image. This can easily be performed in a non-destructive manner utilizing an image processing software program.
The last step of a forensic investigation should be a verification or technical review. A comprehensive and easy-to-follow report should accompany the processing results with all settings and parameters documented in a way that allows for a proper confirmation of the processes performed and, ultimately, the conclusions made by the original analyst.
Any forensic science should meet the general requirements of reliability, repeatability, and reproducibility. Reliability focuses on the idea that the actions performed are known, scientific, documented, and proven to work in the way we expect. With the increase in prevalence of artificial intelligence (AI)- based tools, it is more important than ever to be able to easily reference and prove the reliability of the functions performed on an image. Transparency from a software company in the mathematical algorithms applied to an evidence image is paramount in providing legally defensible processed image evidence.
Repeatability ensures that if the same technique is applied over and over again, it will be implemented similarly. Ideally, a forensic tool would allow you to copy a process onto another image so that all parameters are applied in the same way. This repeatability would allow any similar image to be handled with an identical approach, producing a similar result.
Reproducibility means that another qualified analyst performing the same actions will be able to reproduce similar results. The ultimate goal for the processing of any evidence is to present it to a trier of fact to assist in determining a verdict. An essential part of due process is the ability of an independent opposing expert to analyze an evidence item and generate their own opinion based on analysis, comparison, and evaluation.
Final word
While many commercially available image-processing software programs exist, they are often designed for creative applications rather than forensic use. Because of this, it is challenging to meet the strict requirements of reliability, repeatability, and reproducibility. Forensic image processing demands transparency in the mathematical algorithms applied, as well as clear, structured documentation of every enhancement made to an image. Without these elements, verifying results and ensuring legal admissibility becomes difficult. By utilizing tools specifically designed for forensic applications, analysts can maintain the integrity of digital evidence, streamline their workflow, and facilitate independent verification. This ultimately ensures that forensic image and video evidence hold up to scrutiny in the pursuit of justice through science.
Contact Granite Discovery to learn more about our services and how we can help you.
The original article, “How To: Digital Processing of Forensic Image Evidence,” was written by Melissa Kimbrell, Trainer and Technical Support Specialist at Amped Software. It was posted on the Forensic On the Scene and In the Lab website on March 05, 2025, and can be found here.