The Art of Seeing: How Computer Vision and Human Vision Differ

As humans, we visually perceive the world around us not just in static images but with a combination of colors, textures, motion, depth, and context. On the other hand, computer vision typically focuses on specific visual features without considering anything else. In fact, blocking out as much information as possible that isn’t necessary for the device’s function is essential in computer vision devices to maximize the device’s performance and reliability.

Therefore, one of the most important distinctions when looking at computer vision compared to human vision is which wavelengths the two vision systems are optimized for.

Humans can see light wavelengths from 380 nm, which looks violet to our eyes, to around 780 nm, which appears red to us. This is illustrated in the figure below, where you can also see that the wavelengths shorter than 380 nm are called the UV range, whereas light wavelengths longer than 780 nm are classified as infrared wavelengths.

This distinction is important because computer vision doesn’t necessarily work in quite the same range as human vision. Even though computer vision, in many ways, ‘mimics’ human vision in how it functions, computer vision devices can still be optimized to work in specific wavelength ranges to maximize their performance. Eliminating wavelengths that are irrelevant to the device’s function means that less noise passes through to the sensor, which can hinder its performance.

This is why the signal-to-noise ratio is an important metric in machine vision applications.

Since devices such as displays and HMI panels are optimized for the human eye, the cover glass of these types of applications is designed to maximize the readability in the visible light spectrum. For these devices, features such as scratch resistance, anti-glare or anti-reflective properties, touchscreen functionality, etc., are more important to consider for the performance and longevity of the device.

In many cases, computer vision applications are optimized for infrared wavelengths, aiming to detect or record wavelengths in a specific range above 780nm. This is why we’ve designed optical filters that maximize light transmission in specific near-infrared wavelengths, such as 850nm, in this solution we developed for Cognex.

Solaris™ S306

Solaris™ S306 is one of the most versatile solutions for machine vision applications since it maximizes the transmission of wavelengths above 800 nm up to approximately 2000 nm. This solution is excellent for applications that utilize near-infrared light, such as iris scanners, TOF devices, or cameras designed for ALPR (automatic license plate recognition).

Click here to see how we delivered the perfect solution for AVUTEC’s ALPR devices.

Furthermore, the human eye and brain generate high-resolution images with a wide field of view and excellent sensitivity to different brightness levels.

The cameras in computer vision devices are usually limited by not just a wavelength range, as mentioned above, but the quality and resolution of the components. In some cases, they can also be limited in the speed at which they can process images if the computational power is insufficient for some tasks.

The abovementioned qualities are primarily because human and computer vision have very different jobs. Human vision is a more ‘generalist’ vision system and is not designed for one specific task.

In contrast, computer vision devices are usually designed to do one thing really well, such as facial recognition, scanning barcodes quickly and efficiently, or measuring the distance between two points at any given time.

Computer vision devices are also less prone to errors or mistakes as long as all aspects of the device are well-designed and implemented. This is because good computer vision applications are less prone to optical illusions, biases, and perceptual errors because of fatigue.

In conclusion, human vision has remarkable features and is one of, if not the most important of the five human senses to navigate around in the world. On the other hand, computer vision is usually a highly specialized device at doing one thing consistently and really well.

You wouldn’t trust a computer vision application to make your morning cup of coffee (unless it’s designed to do so), just as you wouldn’t expect a human to detect, read, format, and store a vast number of license plates in a database for future reference.

The proper optical filter or display window is crucial to enhance performance and ensure that the device works as intended. That’s why we are experts at ensuring that the cover glass of your device has the appropriate design and features.

Don’t forget to get in touch with us next time you’re looking for an optical filter or a display window. We would love to help you choose the perfect solution for your project.