A. Akimov, A. Kolesnikov, and P. Fränti (Finland)
: vector map compression, polar quantization, Cartesian quantization, vector quantization, dynamic programming
We consider the quantization problem of lossy vector map compression. The compression is performed by scalar quantization. The scalar quantization is processed in optimal way: we use the Dynamic programming quantization algorithm instead of using uniform or locally optimal Max-Lloyd algorithms. This approach allows us to increase the efficiency of lossy compression for product scalar quantization in rate-distortion sense. We also consider the problem of increasing speed of convergence of existing vector quantization algorithm: randomized local search. The proposed method of using the optimal product scalar codebook as initial codebook increases the speed of convergence for RLS algorithm.
Important Links:
Go Back
