Video Analysis Errors in Murder Trial

A Canadian trial judge convicted the defendant of two counts of murder based on circumstantial evidence and the judge’s own comparison of questioned images of the shooter and known images of the defendant. The Court of Appeal set aside the convictions and substituted them with acquittals due to the trial judge’s flawed analysis of the video recordings and the other evidence in the case. This article discusses the errors that were made and identifies how a lack of visual literacy contributed to this erroneous analysis.

The Trial

Three men in a vehicle were shot by the assailant, two fatally. The central issue before the trial court was the identification of the shooter. The only evidence that linked the defendant to the scene were his fingerprints (4 of 27 fingerprints found) and DNA on a plastic bag that was found on the ground near the shooting and that may have contained the gun prior to the shooting. Expert evidence confirmed that there was no way to determine when the defendant’s fingerprints and DNA were applied to the bag. The gun was never recovered. Other unidentified fingerprints and DNA were also found at the scene and on a vehicle of interest. Commercial CCTV captured relevant parts of the event but was of poor quality.

The trial judge conducted his own comparison of the questioned (CCTV) and known (bodycam arrest) images. He conducted his comparison by pausing at certain frames and taking screen captures. He noted that the CCTV camera was motion activated, the view was somewhat distorted due to a wide-angle view, and that the shooter was pixelated. He found that the quality of the images did not allow for definitive identification. The Crown did not argue that the shooter could be identified from the CCTV images. Upon examining both the CCTV and bodycam images, the trial judge agreed that identification was not possible given the quality of the CCTV images but did note that he found no “dissimilarities” between the questioned shooter and the defendant as depicted. He noted that the shooter and the defendant shared similar size, weight, body shape, facial shape, hairstyle, hairline, back curvature, and facial profile. He found that the combined effect of these similarities increased the probability that both men were the same person.

Based on the fingerprints and DNA on the bag and the judge’s comparison of the video images, the trial judge found that the defendant’s guilt had been proven beyond a reasonable doubt.

The Appeal

The Alberta Court of Appeal (R. v. Benn, 2026 ABCA 14) noted that even though the trial judge acknowledged that the CCTV video was pixelated and of insufficient quality to allow for definitive identification, he nonetheless assigned weight to the similarities that he observed between the two recordings. The Court found that the trial judge engaged in an improper comparison exercise.

[42] The analysis performed by the trial judge carried a high risk of unconscious cognitive bias; specifically, the risk that having seen the high-quality image of the appellant’s arrest, the trial judge would view the low-quality footage of the shooter looking for confirmation that the police had arrested the right person, thereby unconsciously minimizing the possibility that there could be, and probably are, many other people who look similar to both the appellant and the shooter.

[44] The trial judge’s self-instruction on these points addressed some general concerns associated with eyewitness and video identification evidence but did not advert to the specific and somewhat unique danger raised by the evidence in this case: the risk that a trier of fact viewing poor quality video images of an unidentified perpetrator and comparing it to high-quality video image of the prime suspect would unconsciously expect to see significant similarities between the two.

[49] Further, as the trial judge noted, the lens on the CCTV camera distorted the size of objects, making it difficult to precisely discern the height or weight of the shooter and no expert testimony was entered with respect to the likely height and weight of the shooter in the CCTV footage. Combining the potential inaccuracies from a layperson estimating height and weight from distorted video footage with the highly generic nature of the observed similarities, the finding of physical similarity, or of the absence of dissimilarity, provided little to no appropriate support for the ultimate conclusion on identity.

The Court’s criticism of the trial judge’s image analysis and comparison is blunt but apt. Collectively, the Court found that there was a notable lack of sufficient reliable evidence to justify the conviction of the defendant. Accordingly, the Court of Appeal directed an acquittal rather than ordering a new trial. There simply was not enough evidence to ground a conviction.

Commentary

Based on the details set out in the Court of Appeal’s judgment, there was likely insufficient evidence to commence a prosecution, let alone merit a conviction. However, it is the trial judge’s image comparison exercise and the comments of the Court of Appeal that prompted this article. Trial judges (and juries) are required to evaluate all the evidence presented by the parties. This includes examining questioned and known images. The problem is that without expert assistance, judges and juries will do what is natural and contraindicated by science – they will examine the highest quality known images first and then look for correspondence in the questioned images. This process injects cognitive bias and encourages improper findings of fact. The Court of Appeal readily identified this problem. One can hardly fault a trial judge who has been given no expert assistance and is left to sort out image analysis and comparison on his own. A judge in such circumstances is bound to do this exercise backwards.

Had a forensic video analyst conducted a proper image analysis and comparison, it would have been done using scientific methodology. The known images would not even be viewed until the questioned images were fully evaluated and interpreted. Only then would the known images be viewed, evaluated, and interpreted. Then the comparisons and contrast would be embarked upon. A further problem is that there is no evidence that a forensic video technician or analyst processed the images. No reference is made to the images being technically evaluated, adjusted for brightness and contrast, clarified, or in any way improved for comparison purposes. This further hampered the job of the trial judge.

It seems unlikely that a proper forensic video analysis of these images would have made a marked difference given the reported poor quality of the CCTV images, but since that was not done, this is only speculation. More importantly though, when counsel do not retain the services of a properly qualified forensic video analyst, they run the very real risk (realized in this case) of the trial judge conducting their own analysis, which is typically done incorrectly. This is a visual literacy problem. Both counsel and the court need to be alive to the challenges involved with image-based evidence and must seek expert assistance so that fully informed decisions can be made about the value of tendered evidence. Given the seriousness of the offences involved, every technical advantage should be brought to bear in the search for the truth so that a more informed decision can be made about whether there is sufficient evidence to merit a prosecution and if so, to then assist the trial judge (and jury) in their fact-finding role. For more information, see the other articles on this site about image comparison.