bit depth

Bit depth is how many bits are used to store each pixel value, and therefore how many distinct intensity levels are available. With n bits you get 2^n levels: 1 bit gives just black and white, 8 bits give 256 levels per channel, and the brightness steps get finer as the depth grows. It is the resolution of the value axis, the counterpart to spatial resolution which is the resolution of the position axes.

Be careful with the unit. Bits per channel (bpc) describes one color component — 8-bit-per-channel RGB stores 8 bits each for red, green and blue, totaling 24 bits per pixel and 2^24 ~ 16.7 million colors. Bits per pixel (bpp) totals across channels, which is why 8-bit-per-channel color is often loosely called 24-bit color. A single-channel 8-bit grayscale image has 256 levels (0 to 255); 10-bit gives 1024, 12-bit gives 4096, and high-dynamic-range and scientific imaging use 16-bit integers or 32-bit floats to hold a vast intensity range with fine steps.

Bit depth sets the floor on quantization error and the visibility of banding. Too few levels cannot represent smooth gradients, so a gentle sky breaks into visible steps; more bits make the quantization step smaller than the eye or the algorithm can notice. It is the link to quantization (the act of rounding continuous intensity to these levels) and to dynamic range (the ratio of brightest to darkest representable signal). For computer vision and especially HDR, computation is often promoted to floating point to avoid accumulating rounding error, even when capture and display are 8-bit; raw camera files keep 12 to 14 bits precisely so that exposure and tone adjustments do not reveal banding.