Scalar quantization (SQ) refers to any strategy for finding a discrete representation of a continuous scalar variable. The discrete values to which the variable is quantized are called “levels.”
The most conventional approach to SQ is simply to divide the number line into equally spaced intervals, called uniform quantization. A more sophisticated approach involves using k-means clustering along the number line.
In the context of vector quantization, SQ implies quantizing each dimension independently. This can result in an untenably high cardinality, which is why other approaches (such as k-means or product quantization) are preferred.