| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To provide entropy constrained vector quantization for which high coding efficiency can be achieved and which can be easily processed. <P>SOLUTION: The positions of quantization representative vectors are restricted to lattice points existing on the surface of a multidimensional cube (hypercube) with the center at the origin, and the length (2c<SB>i</SB>) of one side of the hypercube having multiple sides is variable. The interval between the lattice points is a value obtained by dividing the length of one side by an integer, wherein the integer is represented by 2K<SB>i</SB>+1, and is also variable. In other words, the arrangement of the quantization representative vectors includes two degrees of freedom; one is the distance from a group of quantization representative vectors to the origin and the other is one fractional number for the distance between each quantization representative vector and the origin. With these two degrees of freedom, the quantization representative vectors are so set as to optimize a cost of coding of the learned data and then vector quantization/reverse quantization is performed. <P>COPYRIGHT: (C)2009,JPO&INPIT</p> |
