< Back to glossary


Quantization is an algorithm for coarsening the coefficients of a discrete-cosine transformation matrix. The human eye pays less attention to small details, which are a high-frequency component. Therefore, during the encoding process, part of the high-frequency spectrum is thrown out. This is done using a quantizer. The frame is divided into blocks, each of which undergoes a discrete-cosine transformation. At the output, we get a matrix where the most significant values are in the upper-left corner, and the least in the lower-right corner. Next, the coefficients of the matrix are divided by the quantization value. Some of the coefficients may become equal to zero. This will give more compression with minimal impact on the image