Disadvantages of Binary Search Codebook Vs Full Search codebook

1)      The reduction in the number of computations in case of binary search comes at a price. The codebook used at the transmitter must store all the intermediate reconstruction levels as well as the final reconstruction levels because the intermediate levels are used in the search.

The codebook size is, therefore increased by a factor of two over the codebook designed by the full search k-means algorithm.

2)      The tree codebook's performance in terms of the average distortion achieved is reduced slightly in typical applications as compared to codebook designed by k-means algorithm.

However, the enormous computational saving more than compensates for the two fold increase in storage requirements and the slight decrease in performance.

Comparison of Vector quantisation Scalar quantisation

The advantage of VQ over SQ lies in its potential to improve performance. The amount of performance improvement possible depends on various factors such as degree of statistical dependence among scalar in the vector.

The performance improvement comes at a price in conputation and storage requirements. Whether or not the performance improvement justifies the additional cost depends on the application.

Vector quantisation is likely to be useful in low bit rate applications, Where any performance improvement is important and the additional cost due to vector quantisation is not too high due to the low bit rate.

Vector quantisation can also be useful in such applications as broadcasting, where the number of receivers is much larger than the number of transmitters and the high cost of transmitter can be to legated. The receiver in vector quantisation has to store the codebook, but it requires only simple table look-up operations to reconstruct a quantised vector.