I'm taking a computer vision/object recognition course. We read a paper that was attempting to classify what "color-word" a human would use to label specific pixels in an image. As I read the paper, I was reminded of some
color-based optical illusions I had seen before. I decided to create some of my own using Emily's colorful teething ring. You can see the result in the image below.
It doesn't look all that remarkable (perhaps a bit Andy Warhol-esque) unless you know what you're looking for. One of the bead colors in each image, is actually not a color at all. It's gray. Perhaps the strongest effect is the yellow beads with the dark blue background. My eyes definitely see yellow, but your computer will definitely tell you that all the pixels in that bead are completely gray. I took the image and used Gimp (my preferred, free, image editing software) to select all pixels of approximately the same color (I just clicked on one of the gray beads). I then colored those pixels white. The result is below.
One of the "green" beads had too much shadow on it to be included completely in the "similar color selection". The point is that the color we perceive is different from the color that arrives at our eyes. It depends on other colors around the image and our perception of lighting and structure of the scene. The question, I guess, is, "How do you teach a computer to do that?"
It doesn't look all that remarkable (perhaps a bit Andy Warhol-esque) unless you know what you're looking for. One of the bead colors in each image, is actually not a color at all. It's gray. Perhaps the strongest effect is the yellow beads with the dark blue background. My eyes definitely see yellow, but your computer will definitely tell you that all the pixels in that bead are completely gray. I took the image and used Gimp (my preferred, free, image editing software) to select all pixels of approximately the same color (I just clicked on one of the gray beads). I then colored those pixels white. The result is below.
One of the "green" beads had too much shadow on it to be included completely in the "similar color selection". The point is that the color we perceive is different from the color that arrives at our eyes. It depends on other colors around the image and our perception of lighting and structure of the scene. The question, I guess, is, "How do you teach a computer to do that?"
2 comments:
Wow, wow, wow! I totally didn't believe you, even with your "proof". So I copied your picture into Irfanview and then cut out samples from each of the so called gray beads...and...you are totally telling the truth! Each of the four samples I took weren't only gray, they were all the same shade of gray. I am amazed!
I wish I could answer your question. All I can say is, Joseph, you are one smarty pants. And, doesn't this make us all way more impressed with the human eye and brain? Thanks for that little reminder to help us remember and appreciate just how much that visual cortex does for us.
Post a Comment