Noise that depends on the subject as much as on the sensor. It arises where fine detail falls awkwardly across pixel boundaries and through the demosaicing arithmetic: a line half in one pixel and half in the next. Because it changes with every scene, the camera cannot mask it out in advance the way it removes hot pixels. Moiré is its most dramatic form.
Faint dark stripes, usually at the edges of objects, that appear when pixel data falls out of step between camera and computer. Of the three kinds of noise it is normally the least serious, and it can often be corrected in editing software.
Specks of purple or white in the dark parts of an image, caused by pixels that switch on slightly early. No two diodes leave the factory identical: each has a faintly different threshold before it responds, and in shadows, where the true signal is tiny, the early switchers stand out. The camera compensates by reading the sensor in total darkness as the shutter is pressed, building a mask of the offenders, and subtracting that mask from the picture.
Shimmering false patterns and colors that appear when a fine, regular pattern in the subject interferes with the regular grid of the sensor. It is a form of random noise: it depends on the subject, so the camera cannot map it out in advance.