Apple Aperture Digital Photography Fundamentals - Manual (Page 43)

Chapter 3 Understanding Resolution 43
Calculating Color and Understanding Floating Point
As you’ve learned, digital devices translate color into numbers. Aperture calculates
color using floating point, a type of calculation that allows calculations to be
performed at a very high resolution with a minimum of error.
Learning About Bit Depth and Quantization
When you capture an image using a digital image sensor, the analog voltage values
have to be converted to digital values that can be processed and then stored. For more
information, see “
Digital Image Sensor ” on page 17. The process of converting an
analog voltage value to a digital value is known as digitization. In the process of
converting an analog voltage value to a digital representation, quantization must be
performed, converting the values to discrete numerical values. The accuracy of each
pixel’s value is determined by the length of the binary word, or bit depth. For example,
a 1-bit binary word can represent only two possible states: 0 or 1. A 1-bit system
cannot capture any subtlety because no matter what the tonal value is, a 1-bit system
can represent it either as 0 or 1 (off or on). A 2-bit binary word can represent four
possible states: 00, 01, 10, or 11. And so on. Most digital RAW image files capture a
minimum of 12 bits per color channel (4096 possible states), allowing for many subtle
degrees of tonal values to be represented. The more bits available for each sample, the
more accurately each color channel’s tonal value can represent the original analog
voltage value.
For example, suppose you use 128 numbers to represent the tonal values of color
channels in each pixel in an image within a range of 1 volt. This means your camera’s
analog-to-digital converter is precise to 1/128 of a volt. Any subtle variations in tonal
values that are more detailed than 1/128 of a volt cannot be represented, and are
rounded to the nearest 1/128 of a volt. These rounding errors are known as quantization
errors. The more the signal is rounded, the worse the quality of the image.