When I snap a picture with my iPhone 4S, the size or resolution of the image is 3264 pixels by 2448 pixels, about 8 million pixels or 8 megapixels. Stored in JPEG format (the most common compression method for digital photographs) every pixel can show one of 256 distinct colors. To allow for 256 colors you need to store 8 positions that can hold either a 0 or a 1. Two bits code for blue, three bits code for green, three bits code for red: R R R G G G B B. For example, 0 0 0 0 0 0 0 0 indicates the color black, 0 0 0 1 1 1 0 0 indicates pure green, because all green pixels take a value of 1.
The mountains I photograph with my iPhone thus require 3264 x 2448 x 8 = 64 million bits or 8 million bytes or 8MB (1 byte = 8 bits).
Let’s take a look at text. Most languages based on the Western alphabet use between 20 to 40 letters, plus 10 numbers and some punctuation marks. 64 characters is sufficient to account for most pieces of text. For every character – for example, the letter “a”, 6 bits (2^6 = 64) should be sufficient. If an average word has 6 letters, one word requires 36 bits (6 letters x 6 bits/letter) or 4.5 bytes. One thousand words equal approximately 4500 bytes. (This is 4.5kb. Did you ever notice when saving a text file how little computer memory it uses compared to storing a picture?).
The memory required to store a typical picture can also be used to store 1.7 million words (8 x 10^6 bytes / 4.5 bytes per word)!
[edit: Jeroen Offerijns corrected me saying that computers use 8 bits/letter because of encoding standards. 8 bits equal 1 byte, meaning one-thousand words with on average 6 letters per word require 6,000 bytes of memory. Thus, one iPhone 4s picture equals ~1.3M words (8 x 10^6 bytes / 6 bytes per word).]
Our brains have evolved to capture the richness of a picture practically instantly – we do not disaggregate a picture of a parrot into a matrix of 0s and 1s. If our goal is to minimize time spent on learning or amusement, eyeing at a picture is more efficient than reading a thousand words.
For digital storage, the story is different. It is much more efficient to store millions of words (equaling 10s of MBs) than tens of high-quality images (equaling 100s of MBs); the millions of words likely provide more information too. If uploading information to our augmented brains is part of our future – words will be the clear winner over images.
Welcome to the beautiful field of Information Theory! The seminal paper, and still an awesome introduction into the field, is Claude E. Shannon’s “A Mathematical Theory of Communication” (1948). The theory he presents there is still the basis behind error correction, compression and machine translation.
Remco, thanks for this recommendation.
I read about Claude Shannon’s work not in academic papers but in “Idea Factory” by Jon Gertner, a review of which you find here: http://www.nytimes.com/2012/04/08/books/review/the-idea-factory-by-jon-gertner.html?pagewanted=all&_r=0