# Statistical Reasoning of Noise Images on Random Pixel Generator

http://www.pixelmonkeys.org/#theory

It is always explained that even billions of images are generated per second, it is almost impossible to see a natural image ( whether it is clear or distorted as such picture of Eiffel Tower, picture of Eiffel Tower with a lot of black pixels ) on a random pixel generator during entire lifetime of universe. Rather it is always concluded that noise images like an untuned TV screen are only we can get because of the fact that total number of natural images constitute extremely small percentage among all displays.

Though it is easy to write down number of all possible images a random pixel generator can display, it isn't clear cut to calculate number of all possible natural images.

How do you then statistically conclude vast majority of displays from a high resolution random pixel generator make noise images rather than natural images?

Is it technically based on the following reasoning? Let’s take 800 x 600 screen, each pixels having 16 different colors. Let’s be absurdly generous to define total quantity of natural images that can be produced and say 15 colors out of 16 on each pixel generates a natural image, assigning highest number of colors so that we can get noise images as well. This makes 15^800*600 natural images, an incredibly huge number. However all possible images makes 16^800x600, even way more huge number. When we compare number of natural images with all possible images random pixel generator can produce it still has significantly low percentage. Therefore noise images vastly outnumbers the natural ones.