hrvatski

Vektorska kvantizacija s klasifikacijom rubova i primjena na inverzno polutoniranje

In this dissertation a new algorithm for low bit rates inverse halftoning has been proposed. This algorithm integrates the processes of inverse halftoning and compression (vector quantization). In order to efficiently implement algorithm, a new method for classified vector quantization memory reduction for has been proposed. In addition to the memory reduction, this method gains signal/noise ratio so the method has been applied to proposed inverse halftoning algorithm. Three variants of the algorithm has been proposed: Method A, Method B and Method C. Method A includes table look-up and vector quantization. Method B is an extension of Method A with memory reduction method. Finally Method C is an extension of Method B with edge classification, which means that Method C includes table look-up, memory reduction method and classified vector quantization. The proposed method has been evaluated with three algorithms for image halftoning: error diffusion, dispersed dither and clustered dither. Since these three algorithms are the most frequently used ones, it means that the proposed algorithm has been evaluated for almost all real situations. Experimental results of the proposed algorithm and comparison with similar methods, shows that, for the low bit rates (up to 0.5bpp), that the proposed algorithm is better than the other methods. Also, it has been shown that with the proposed algorithm, the halftoned image can be compressed significantly in gray scale domain, with only slightly lower quality.

vektorska kvantizacija; klasifikacija rubova; inverzno polutoniranje

nije evidentirano