Hypercomplex Cells: How Does Visual Processing Work in Our Brains?

Something super powerful I’ve come across in visual science– the notion of a ‘hypercomplex cell‘:

1. Response to movement with an end point

One interesting takeaway:

  1. If you have a visual graphic without an endpoint (a line that goes straight through the picture), your brain doesn’t give a visual response.
  2. If you have an end-point in your picture, your brain notices it.

In other words:

If you want to make a photograph that catches the attention of your viewer, have an ‘end point’ in your picture.


Let me explain some more. Let us assume you have a blank canvas:

1. No response (no end point):

If you have a line through your canvas which is uninterrupted, and flows from edge to edge, your brain will see some sort of “harmony”. Your brain neurons won’t fire. In other words, you don’t really pay much attention to this image:

2. End-point (within a picture):

However, what if the line isn’t allowed to extend until the edge of the picture? Then you kind of get visual “blue balls”— you want to see the line extended until the edge, but you don’t! There is an end-point within the picture, which catches our attention:

You see the point, and your brain visual processing neurons fire.


3. What if you have a bright dot in the center of a dark background?

Another interesting idea:

If you have a bright (yellow) dot in the middle of a dark (purple) background, your ‘retinal ganglion cell‘ will fire:

That means, the yellow dot in the center of the frame catches your attention!

Theory:

You are more likely to notice a bright spot against a dark background, rather than a dark spot against a bright background.

4. Dark dot inside a bright background

When you have light on the surrounding background, and a dark spot in the center– your ‘retinal ganglion cell’ doesn’t fire.

Takeaway point:

If you’re shooting portraits of people, it is best to photograph your subject against a black background:

5. Example: Portrait against black background

For example this picture Benjamin Thompson shot of me, he just photographed me against a simple black wall:

ERIC KIM x HENRI NECK STRAP // Photo by Benjamin Thompson

With Gaussian Blur, and you can see the separation of me against the dark background:

Now I used Photoshop with the ‘polygon lasso’, used the eyedrop tool (I) to select color regions, and painted them in:

Now totally darkening the background. You can see there is good separation (figure to ground) between me and the background:

Lesson: If you want to make a stronger portrait of your subject, start off with a black canvas (black background), and then have your subject stand in front of it. To further create separation between them and the background, you can shoot with a flash.


6. Example: Subjects against bright background:

Here are some examples of portraits of subjects against bright backgrounds. Let us start with Cindy, shot on RICOH GR II, with flash, in P mode, against a white background:

 

It is my general observation that it is more difficult to make out, or to separate the subject against a bright or white background.

For example, let us add gaussian blur:

Then let us fill in the details:

Then fill the background with the slight off-grey of the real background:

And now, with a pure white background:

A bit difficult to see the central subject (Cindy).

Now what I will do is flatten the layers, and now inverse the whole image:

Instantly I think Cindy (as a subject) ‘pops’ out more from the background.

Now this is when we inverse the initial grey background:

Now aesthetically, I actually prefer the bright white against the slate-grey background.

Lesson: Photographs of people against dark/black backgrounds work best! But still better to photograph a subject against a white background, than a messy background.

For good examples of epic portraits against a simple white background, study the work of Richard Avedon.


Conclusion

Never stop experimenting.

I recommend studying machine vision learning (I recommend the Stanford Online CS231n: Convolutional Neural Networks for Visual Recognition) course, or learn more about their vision lab.

Distort your images, rotate them, flip them, blur them, etc. This will help you better understand your visual images:

Never stop experimenting!

ERIC


Machine Learning + Vision

Brave new world of photography:

Scroll to Top