Abstract
In the subband image coding, the optimum band partition method is to minimize the quantization noise power in the encoded image when the coding rate is given. However, the previous method has the problem of the computational costs to determine the optimum band partition in the 2-dimensional frequency domain. In order to solve the problem, the new algorithm for the optimum band partition is proposed by giving the attention to the kurtosis of the subband signals. As the results, our algorithm can improve the computational costs about 1/75 than the previous method.
