Method and apparatus for low frequency removal in vector quantization
First Claim
1. In a vector quantization data compression system employing a first vector quantization codebook for compressing an image frame for transmission to a plurality of reception sites, said codebook having a plurality of codevectors, each codevector being representative of a possible residual vector, there being an identification (ID) code associated with each codevector, a method comprising the steps of:
- a) receiving an image frame to be transmitted, organizing the image frame into a plurality of blocks of data, and converting each block to a multi-dimensional input vector;
b) determining a mean value of each input vector and transmitting at least an indication of each mean value to at least one reception site;
c) performing the following additional steps for each input vector;
i) processing at least a portion of said means values to produce a low-frequency component vector representative of a low-pass version of the input vector;
ii) subtracting the low-frequency component vector produced in step (c)(i) from the input vector to obtain a low-frequency removed vector;
iii) determining a mean value of the low-frequency removed vector and subtracting the mean value of the low-frequency removed vector therefrom to obtain a residual vector;
iv) selecting from the first codebook the codevector that most closely resembles the residual vector; and
v) transmitting at least an indication of the ID code associated with the selected codevector to at least one reception site, said indication representing vector quantized data.
3 Assignments
0 Petitions
Accused Products
Abstract
A method and apparatus for low-frequency removal in vector quantization of an image frame. An image frame is converted to plurality of input vectors. A mean value of each input vector is determined and transmitted to a plurality of reception sites. A low-frequency component of each input vector is then determined and removed from the input vector. A scalar mean value of each low-frequency removed vector is then determined and subtracted therefrom to produce a residual vector. The residual vectors are then vector quantized and the vector quantized data is transmitted to the reception sites. At a reception site, the transmitted mean values and vector quantized data are received and each input vector is reproduced therefrom. From the reproduced input vectors, the image frame is recreated for display at the reception site.
-
Citations
46 Claims
-
1. In a vector quantization data compression system employing a first vector quantization codebook for compressing an image frame for transmission to a plurality of reception sites, said codebook having a plurality of codevectors, each codevector being representative of a possible residual vector, there being an identification (ID) code associated with each codevector, a method comprising the steps of:
-
a) receiving an image frame to be transmitted, organizing the image frame into a plurality of blocks of data, and converting each block to a multi-dimensional input vector; b) determining a mean value of each input vector and transmitting at least an indication of each mean value to at least one reception site; c) performing the following additional steps for each input vector; i) processing at least a portion of said means values to produce a low-frequency component vector representative of a low-pass version of the input vector; ii) subtracting the low-frequency component vector produced in step (c)(i) from the input vector to obtain a low-frequency removed vector; iii) determining a mean value of the low-frequency removed vector and subtracting the mean value of the low-frequency removed vector therefrom to obtain a residual vector; iv) selecting from the first codebook the codevector that most closely resembles the residual vector; and v) transmitting at least an indication of the ID code associated with the selected codevector to at least one reception site, said indication representing vector quantized data. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14)
-
-
15. In a vector quantization data compression system employing a first vector quantization codebook for compressing an image frame for transmission to a plurality of reception sites, said codebook having a plurality of codevectors, each codevector being representative of a possible residual vector, there being an identification (ID) code associated with each codevector, a method comprising the steps of:
-
a) receiving an image frame to be transmitted, organizing the image frame into a plurality of blocks of data, and converting each block to a multi-dimensional input vector; b) determining a mean value of each input and transmitting each mean value to the reception sites; c) performing the following additional steps for each input vector; i) processing at least a portion of said mean values to produce a low-frequency component vector representative of a low-pass version of the input vector; ii) obtaining a measure of difference between the low-frequency component vector produced in step (i) and the input vector; iii) transmitting an instruction to the reception sites to indicate that no further data will be transmitted for the input vector when the measure of difference is less than a threshold value, but otherwise performing the following additional steps (iv) through (vii) for the input vector; iv) subtracting the low-frequency component vector produced in step (c) (i) from the input vector to obtain a low-frequency removed vector; v) determining a mean value of the low-frequency removed vector and subtracting the mean value of the low-frequency removed vector therefrom to obtain a residual vector; vi) selecting from the first codebook the codevector that most closely resembles the residual vector; and vii) transmitting at least an indication of the ID code associated with the selected codevector to the reception sites;
the indication of the ID code representing vector quantized data;whereby vector quantized data for the input vector is not transmitted to the reception sites when the measure of difference obtained in step (c) (ii) is less than the threshold value. - View Dependent Claims (16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31)
-
-
32. Image compression apparatus comprising:
-
means for receiving a plurality of multi-dimensional input vectors representative of an image to be compressed; a first mean value calculator for calculating a scalar mean value of each input vector; an interpolating low-pass filter for interpolating a low-frequency component vector representative of a low-pass version of a current input vector, the low-frequency component vector being interpolated from the scalar mean values of at least a portion of the input vectors; a first subtracting circuit for subtracting the low-frequency component vector from the current input vector to produce a low-frequency removed vector; a second mean value calculator for calculating the scalar mean value of the low-frequency removed vector; a second subtracting circuit for subtracting the scalar mean value of the low-frequency removed vector therefrom to produce a residual vector; a codebook having a plurality of codevectors, each codevector being representative of a possible residual vector, each codevector having an associated identification (ID) code; a controller for selecting one of the codevectors in the codebook that most closely resembles the residual vector; and a transmitter for transmitting data indicative of the ID code of the selected codevector. - View Dependent Claims (33, 34, 35, 36, 37, 38)
-
-
39. A decoder for use at a reception site of vector quantization data, wherein the vector quantization data originates from an encoding site employing a vector quantization encoder to compress input vectors indicative of blocks of data defining an image frame, the encoder transmitting a scalar mean value of each input vector and additionally transmitting, for each input vector, data indicative of an identification (ID) code associated with a residual codevector selected from a first vector quantization codebook, the decoder comprising:
-
a second codebook being substantially identical in content to the first codebook; a first memory for storing at least the scalar mean values, transmitted from the encoding site, of input vectors for a current image frame; a second memory for storing the ID code transmitted from the encoding site for each input vector of the current image frame; an interpolating low-pass filter coupled to the first memory for interpolating a low-frequency component vector representative of a low-pass version of an input vector to be reconstructed and for providing the low-frequency component vector on an output thereof, the low-frequency component vector being interpolated from the scalar mean value of the input vector to be reconstructed and the scalar mean values of adjacent input vectors; a mean value calculator for calculating the scalar mean value of the low-frequency component vector and for providing the scalar mean value of the low-frequency component vector on an output thereof; a subtractor circuit having a first input coupled to the output of the interpolating low-pass filter and a second input coupled to the output of the mean value calculator for subtracting the scalar mean value of the low-frequency component vector therefrom to produce a zero-mean low-frequency component vector and for providing the zero-mean low-frequency component vector on an output thereof; a controller for accessing the second memory to obtain the ID code transmitted for the input vector to be reconstructed and for retrieving from the second codebook, the residual codevector having the ID code associated therewith and for providing the residual codevector on an output of the controller; a summing circuit having a first input coupled to the output of the subtractor circuit, a second input coupled to the first memory, and a third input coupled to the input of the controller for adding the zero-mean low-frequency component vector for the input vector to be reconstructed, the scalar mean value stored in the second memory for the input vector to be reconstructed, and the residual codevector retrieved from the second codebook for the input vector to be reconstructed, and for providing the sum on an output thereof, the sum being a substantial representation of the input vector; at least one converter coupled to the output of the summing circuit for converting the substantial representation of the input vector to NTSC format image data and for providing the NTSC format image data at an output thereof for display on a display device at the locale of the reception site. - View Dependent Claims (40, 41, 42)
-
-
43. A decoder for use at a reception site of vector quantization data, wherein the vector quantization data originates from an encoding site employing a vector quantization encoder to compress input vector indicative of blocks of data defining an image frame, the encoder transmitting a scalar mean value of each input vector and additionally transmitting, for each input vector, either (i) data indicative of an identification (ID) code associated with a residual codevector selected from a first vector quantization codebook, or (ii) an instruction to approximate the input vector by a low-frequency component vector interpolated from the scalar mean value of the input vector and the scalar mean values of adjacent input vectors, the decoder comprising:
-
a second codebook being substantially identical in content to the first codebook; a first memory for storing at least the scalar mean values, transmitted from the encoding site, of input vectors for a current image frame; a second memory for storing (i) the ID code or (ii) the instruction transmitted from the encoding site for each input vector of the current image frame; an interpolating low-pass filter coupled to the first memory for interpolating a low-frequency component vector representative of a low-pass version of an input version to be reconstructed and for providing the low-frequency component vector on an output thereof, the low-frequency component vector being interpolated from the scalar mean value of the input vector to be reconstructed and the scalar mean values of adjacent input vectors; a mean value calculator for calculating the scalar mean value of the low-frequency component vector and for providing the scalar mean value of the low-frequency component vector on an output thereof; a subtractor circuit having a first input coupled to the output of the interpolating low-pass filter and a second input coupled to the output of the mean value calculator for subtracting the scalar mean value of the low-frequency component vector therefrom to produce a zero-mean low-frequency component vector and for providing the zero-mean low-frequency component vector on an output thereof; a controller for accessing the second memory to determine whether an ID code or an instruction was transmitted for the input vector to be reconstructed, and where an ID code was transmitted for the input vector to be reconstructed, retrieving from the second codebook, the residual codevector having the ID code associated therewith and for providing the residual codevector on a first output thereof and a control signal on a second output thereof; a summing circuit having a first input coupled to the output of the subtractor circuit, a second input coupled to the second memory, and a third input coupled to the first output of the controller for adding the zero-mean low-frequency component vector for the input vector to be reconstructed, the scalar mean value stored in the second memory for the input vector to be reconstructed, and the residual codevector retrieved from the second codebook for the input vector to be reconstructed, and for providing the sum on an output thereof, the sum being a substantial representation of the input vector; a multiplexor having first and second data inputs and a data selector input, the first data input being coupled to the output of the summing circuit, the second data input being coupled to the output of the interpolating low-pass filter, and the data selector input being coupled to the second output of the controller for providing on an output of the multiplexor, in accordance with the control signal from the controller, either (i) the substantial representation of the input vector, or (ii) the low-frequency component vector which represents a low-pass version of the input vector; and at least one converter coupled to the output of the multiplexor for converting the multiplexor output to NTSC format image data and for providing the NTSC format image data at an output thereof for substantially reproducing the image frame that was input to the encoding site, whereby if the controller determines that an ID code was transmitted for the input vector to be reconstructed, then the controller selects the first data input of the multiplexor and the substantial representation of the input vector is converted to NTSC format image data, but if the controller determines that an instruction was transmitted for the input vector to be reconstructed then the controller selects the second data input of the multiplexor and the low-frequency component vector, which represents a low-pass version of the input vector, is converted to NTSC format image data as an approximation of the input vector. - View Dependent Claims (44, 45, 46)
-
Specification