Vector transform coder with multi-layered codebooks and dynamic bit allocation
First Claim
Patent Images
1. A method of producing a multi-layered codebook structure for image coding, comprising the steps of:
- transforming a plurality of training images into respective vector components in the vector transform domain;
calculating a first centroid value for each vector component;
calculating a second centroid value for each vector component by subtracting a mean value of said vector component therefrom;
deriving a first codebook layer having a plurality of codebook elements by executing a training algorithm using said vector components and the first centroid values;
deriving a second codebook layer having a plurality of codebook elements by executing a training algorithm using said vector components and the second centroid values; and
deriving a third codebook layer having a plurality of codebook elements by scaler quantization of said vector components.
3 Assignments
0 Petitions
Accused Products
Abstract
In a vector transform coding (VTC) scheme for image and video compression, a multi-layered codebook structure and a dynamic bit-allocation scheme in the vector transform domain are employed. The main advantage of this technique is that distortion is controlled by dynamically allocating more bits to vectors causing larger distortions and less bits to vectors causing smaller distortions.
27 Citations
1 Claim
-
1. A method of producing a multi-layered codebook structure for image coding, comprising the steps of:
-
transforming a plurality of training images into respective vector components in the vector transform domain; calculating a first centroid value for each vector component; calculating a second centroid value for each vector component by subtracting a mean value of said vector component therefrom; deriving a first codebook layer having a plurality of codebook elements by executing a training algorithm using said vector components and the first centroid values; deriving a second codebook layer having a plurality of codebook elements by executing a training algorithm using said vector components and the second centroid values; and deriving a third codebook layer having a plurality of codebook elements by scaler quantization of said vector components.
-
Specification