Following the considerable success that linearpredictivecoding (LPC) has had in speech compression, the technique has been applied to the coding of two-dimensional (2-D) signals such as natural images. Unlike its on...
详细信息
Following the considerable success that linearpredictivecoding (LPC) has had in speech compression, the technique has been applied to the coding of two-dimensional (2-D) signals such as natural images. Unlike its one-dimensional (1-D) counterpart, the 2-D technique is not guaranteed to be stable. It is found that too much correlation in the signal causes a significant proportion of the analysis frames to produce unstable prediction filters, rendering the decoded image unintelligible. The paper introduces two methods for systematically reducing the signal correlation, and hence improving the so called ‘stability rate’ of a 2-D LPC system. The first method is based on the 2-D Fourier transform, and the second is based on the 2-D Hadamard transform. The effectiveness of each method is illustrated followed by a cost analysis based on algorithm complexity and bit-rate overhead.
暂无评论