Vector quantization method employing mirrored input vectors to search codebook

DETAILED DESCRIPTION Briefly stated, the present invention is directed to a vector quantization method that employs mirrored input vectors to increase the reproduction quality of transmitted vector quantized data.
In the preferred embodiment, the method is embodied in a pay television system wherein image data (usually program material, such as movies, etc.
) is vector quantized at a transmitter location of pay television signals for transmission via satellite to a plurality of receiving locations.
The image data typically comprises temporally spaced image frames representing moving images to be displayed on a television set at each receiving location.
The method comprises first receiving an image frame to be transmitted, organizing the image frame into blocks of image frame data, and converting each block of image data to a multi-dimensional input vector.
Each input vector has a plurality of possible orientations.
An identifying code is assigned to each orientation.
A first vector quantization codebook having a plurality of codevectors is provided in a memory at the transmitter location.
The codevectors are representative of possible input vectors.
Additionally, each codevector has an associated memory address.
For each input vector, the following steps are performed.
First, one of the orientations of the input vector is selected.
The input vector with the selected orientation is then compared to the codevectors in the codebook, and the codevector that most closely resembles the input vector (in the selected orientation) is selected from the codebook.
A measure of difference is then obtained between the input vector with the selected orientation and the selected codevector.
A different one of the remaining orientations of the input vector is then selected, i.
e.
, the input vector is re-oriented such that it has a different one of its possible orientations.
The input vector, in its newly selected orientation, is again compared to the codevectors in the codebook and the codevector that most closely resembles the input vector (in its newly selected orientation) is selected