January 24, 2016

A Battle for the red light !

How well do you see the differences between two colors intensities? Can you distinguish them better than a machine can? That’s what our team decided to discover this week, and this blogpost will show you how we did it:

But first, let’s try and understand what colors are, and how it is that we see them.
Colors are waves that travel through space, with a specific frequency. We can see because our eyes have two different kinds of sensor: cones and rods. Rods help us see when the luminosity is very low. They are the reason why we see in the dark. Cones are the reason we see colors. They are sensitive to specific frequencies, and are sensible to red, green and blue. So how can we see yellow, purple and white? Well, yellow is a bit of red and a bit of green; purple is a bit of red and a bit of blue, while white is made of all three colors at the same time.
This is also how the colors of our computer screens are made. Pixels are made of Red, Green and Blue in varying intensities, and the intensity of each color determines the overall color we see.
With intensities varying from 0 to 250 for green, blue and red, there are 16,777,216 possible color values, but can we distinguish them all?

DSC_0308.JPG
Figure 1: in this scheme of an human eye you can see the emplacement of cones and rods
What we tested was simple enough: we coded a short program that consisted in randomly alternating between two given colors. We selected two shades of red that were really close in intensity and ran them through the program. We then asked people if they could distinguish between the two colors or if they did not see a difference. By repeating this experiment with a more noticeable difference, we aimed to determine the resolution limit for the human eye. We then compared that with the limit of resolution of an electronic sensor.

To do that, we repeated the same protocol but this time, instead of a person watching, it was an Light Dependant Resistor (LDR) device. An LDR device is a resistor that has a different resistance value depending on how much light it is exposed to. Since the intensity of the color involves a difference in the light the sensor perceives, we were able to determine the sensitivity of the device and compare it to the sensitivity of our eyes.
Figure 2: by looking at this image, you only notice the contrasts between the last rectangle and the rest. However, there are 20 different nuances of red your eyes cannot distinguish one to one!

We therefore interviewed more than 30 people and ran the program in front of them. We noted how many of them noticed each difference of color, and this allowed us to estimate the threshold of human perceptions.
We did the same thing for our intensity-sensing device and plotted a graph of the data it fed us back. The sensor consisted in an Light Dependant Resistance, plugged into a small computer chip called an Arduino, and connected to a LED, that lit up every time the sensor caught a difference between the two colors. The point was to determine the device’s sensitivity compared the human eye. After that, we determined which of these 2 sensors was the most efficient to detect small light intensity changes for red color. The tricky part was to calibrate the electronic sensor so that it would spot the finest difference, but not react on what was just some variations of the environment.

So what did we find? Well, the average difference of nuances a human needs to spot a difference between two intensities of red is of 4,25. the minimum for a machine is… One.
Of course, this varies on individual perceptions, for instance, some women have an additional kind of cone that help them see a lot more color than average people!

Untitled Infographic(2).jpeg
Figure 4: with this infographic you can observe percentages of people who saw a difference of light intensity and which value of difference.

In conclusion, if you were to look at the 250 shades of pure red we can code with pixels, you would see about 60 if your eyes are average, and a well-calibrated LDR sensor would see all of them. Incredible, isn’t it?

For further studies, look at the following links:
Video about differences of color perception from one person to another:  

Authors of the project : Anne-Pia MARTY, Hortense VIELFAURE, Nathan TOUBOUL and Louis GALLARD

No comments:

Post a Comment

Blog disclaimer

The content created by the Learning thru research Student Bloggers and those providing comments are theirs alone, and do not reflect the opinions of Centre de recherche interdisciplinaire, University Paris Descartes or any employee thereof. The authors of posts and comments are responsible for the accuracy of any of the information supplied on this blog, as well as for any content copyright issues.