Scalar quantization forces source values into some fixed subset of quantized values. For example, say we have a quantization table which allows only the values 0 and 1. If we choose to quantize to the closest value, then .4 would quantize to 0 and .8 would quantize to 1. Of course, the larger the number of possible values in the quantization table, the more precise the output from the quantization will be to the original.
Vector quantization quantizes arrays of values rather than a single value.
TakeDown.NET -> “Quantization”