Conceptualizing Color in a Binary Image
Or Communication as arbitrary assignment
Color is everywhere, forming frightfully easily identifiable separations used to categorize, organize, and define. Color is so naturally interpreted with such consistently among humans that its application requires no additional explanation. It is simply experienced.
In deconstructing and reconstructing the experience of color, I will attempt to develop an understanding of symbolic generation.
The problem: How to best select a general purpose static codec that translates 24-bit color images into 1-bit monochomatic images:
The objective of the codec is not to optimize for the interpretability of objects in the 1-bit monochromic image for immediate human consumption but rather to optimize the decoding of 1-bit monochromic images back into 24-bit color images.
Basically, design an algorithm to map each 24-bit color value to a 1-bit value which provides the highest probability of being decoded back into a 24-bit color value.
Assumption 1: The best possible mapping of values is a random selection
Assumption 2: A biased dataset (order, patterns) crossed with an unbiased dataset (chaos, randomness) results in a new biased dataset (order, patterns)
Deconstructing from 24-bit color images to 1-bit monochromatic images using a randomly selected mapping of color codes:
Figures 3–8 empirically and convincingly confirm assumptions 1 and 2. Figures 9–12 empirically confirm as well only with less conviction.
Constructing generated 1-bit monochromatic images back into 24-bit color images using symbolic generation:
Assumption 3: There must exist distinguishable biases (order, patterns) in a dataset which occur above a threshold as to be experienced and warrant the assignment of a symbol (learning)
Algorithm:
1. Feed x number of 24-bit images through the codec A to generate 1-bit monochromatic images
2. Apply learning to identify S1 symbols in the 1-bit monochromatic images
3a. If the number of S1 symbols is less than the number of S24 symbols in the 24-bit color images (16,777,216), repeat step 1
3b. Else randomly select S1 and apply them 1–1 to the S24 symbols to generate new codec B
4. Replace codec A with codec B
As a result, the 1-bit monochromatic observer is able to encode with greater precision and variety and, theoretically, increase the amount of information and meaning conveyed to the 24-bit color observer.
An Application Towards Human Communication
Applying the same process to the development and exchange of language, humans are able to convey meaning without actually experiencing the conveyed meaning of an interlocutor so long as the underlying experiences are distinguishably similar.
Some other thoughts to observe are the emergence of rhyming, idioms, refinement, and taste. Take for instance the 1-bit monochromatic and 24-bit color image observer of the world. I hypothesize that patterns will emerge over innumerable interactions that only a 1-bit monochromatic observer will experience.
If you are interested in critiquing and collaborating, the GitHub repository is available for pull requests and the database with HTML color code mappings for your consumption.
