Seeing Colors – The Inverse Problem and Big Data Analysis

This is my final essay for The Visual Perception and the Brain on Coursera. I combined my interest with my expertise by explaining how our visual perception works in the concept of big data analysis. I’ve got full points from 5 peers in the end…though some of them were overwhelmed by my crazy idea 😛 Still, it’s awesome!!

I’m always fascinated by colors since I was young, and I wondered why red is red, and blue is blue. Even though I knew well that the colors we see are perceptual properties derived from the visible range of the spectrum of light, why we see what we see remains a puzzle.

Not until I took this course, did I realize color is not an absolute perception but a relative interpretation. We see something as red not because the surface has reflected a certain wavelength, like 700nm, of light, but by comparison of the signals our cone cells perceive.

By the definition of the inverse problem, the image formed on the retina is a conflation of information of the physical world, and we have no way to extract real-world information from this image. Therefore, we could deduce we don’t see color according to its physical property, but from other data source – experience, as concluded throughout the lecture.

This idea can be further consolidated by Color Constancy. Why are we able to identify the red of an apple whether it’s under the sunlight, fluorescent lighting, or neon lighting, when it seems obviously not red? The brain is able to discount the effect of the continual change in the wavelength composition of the illuminating light reflected form a surface. This stability in color as we perceive it allows the brain to obtain knowledge about the properties of surfaces and recognize the “true color”, despite continual variations in what reaches the eye from those surfaces.[i] [ii]

color consistency

Color is a property of objects that our mind creates; an interpretation of cues that our visual brain perceives. We have been taught or trained to identify the “true color” of a certain object whether which condition it’s in, ex: under the influence of different types of light and degrees of shade, since we were born. We therefore associated environmental cues with the color we saw, and preserved this combined information into our “brain database”, as reference that helps to discard the illumination influence.

Thus, we could say the perception of color is the result of a learning process of our brain and visual cortex based on prior memories and experiences.[iii] An adult who restores his/her vision from blindness caused at birth or in childhood might not be able to identify the color of an object or even tell the difference among colors in the first place[iv], due to the lack of proper visual stimulation in early age, which prevented him/her from developing the physiological stage of visual processing. The optical stage provides the raw message, but it is the physiological stage that determines what can be seen.[v]

In the same vain, we can further explain how experience, our “brain database”, has helped us to bypass the inverse problem and developed workable visual interpretation that can interact with the world successfully.

Recently people have been talking about big data analysis[vi], and actually this is the manifestation of how human deals with the inverse problem[vii] in color perception. In big data analysis, we collect huge amount of field data (all the physical world we perceive visually), classify, analyze and extract feature from it (ex: considering different types of illumination as a data parameter). Then we build and develop a workable model to describe the relationship between the data parameters and the outcome, which, in terms of color vision, is an analogy of how human is able to identify the influence of different illumination conditions and discount its effect, so as to obtain an object’s reflectance – the wavelengths of light the object reflects, and identify its “true color”.[viii]

Big Data

Human brain is much more complicated and sophisticated than any computers we’ve ever built in the world, and long before we knew how to utilize data mining to help with decision making, our brain has developed a mechanism to deal with the big data collected from prior experience and “auto correct” our color vision regardless of the lack of real information of the physical world that forms the inverse problem.

However, sometimes this mechanism causes “over-correction” and hence the Color Contrast example proposed in the Lecture 5. In the illustration below, picture (B) is an example of Color Constancy, which demonstrates the brain is able to perceive stability in color despite the influence of illumination or shade. However, since our brain tends to dismiss the effect of illumination composition, when the perceived wavelength of the light reflected by two areas are actually the same, sometimes the non-existent illumination and shade influence are mistakenly taken into account and “discounted”.

color contrast

Just like picture (A) above, the darker color tone of the front facet suggests dimmer light source, which tricks the brain to “brighten” the color perception of the whole front facet, and leads us to view the orange square in the middle as brighter than the brown square in the middle of the top surface.

Nevertheless, why we see what we see is the result of human evolution, derived from the experiences of coping with the world successfully. No model is perfect, whether in big data analysis or in human color vision. We perceive this “strange” phenomenon because there’s no sufficient data collected in our “brain database” and served as valid input to adjust the model yet, just like the ancient people had difficulties to identify blue and distinguish blue from green[ix]. In time, I believe the human color vision will take a step forward again, become more sensitive and adapt to the physical world more precisely.

[iii] Cheryl Kamei Hannan (2006), Review of Research: Neuroscience and the Impact of Brain Plasticity on Braille, Journal of Visual Impairment & Blindness, v100 n7 p397-413

[Appendix – comments from my peers]

peer 1 The parallel made between perception and big data analytics seems a bit far fetched but non the less it makes sense. The essay also makes clear references to the contents of the course.
peer 2 Good work. Very clearly explained, cited resources with were apt for the discussion. Overall I liked your presentation of the topic. Good job.All the best!
peer 3 The author in detail reviews the inverse problem and how our visual system associated with color perception cope with it. The essay also contain 3 figures which support her examples and better illustrate the subject. Well done!
peer 4 You began with a strong thesis statement and did a great job of keeping the inverse problem in the forefront throughout the paper. The diagrams and illustrations you chose were dynamic to your subject matter. Demonstrated good understanding of the course material. Excellent.
peer 5 The student wrote about a more specific feature of the visual system but in a very clear way. He wrote a very concise and specific essay.
本篇發表於 Coursera 並標籤為 , , , , , , , , , , , 。將永久鏈結加入書籤。


在下方填入你的資料或按右方圖示以社群網站登入: 標誌

您的留言將使用 帳號。 登出 /  變更 )

Google+ photo

您的留言將使用 Google+ 帳號。 登出 /  變更 )

Twitter picture

您的留言將使用 Twitter 帳號。 登出 /  變更 )


您的留言將使用 Facebook 帳號。 登出 /  變更 )


連結到 %s