Adaptive block transform image coding method and apparatus

• US 4,774,574 A
• Filed: 06/02/1987
• Issued: 09/27/1988
• Est. Priority Date: 06/02/1987
• Status: Expired due to Term

• Current Assignee: Eastman Kodak Company
• Original Assignee: Eastman Kodak Company
• Inventors: Daly, Scott J., Rabbani, Majid, Chen, Cheng-Tie
• Primary Examiner(s): Britton, Howard W.
• Application Number:
  US07/057,404
• Time in Patent Office: 483 Days
• Field of Search: 358/133, 358/138
• US Class Current: 348/406.1
• CPC Class Codes: H04N 19/124 Quantisation
• CPC Class Codes: H04N 19/126 Details of normalisation or weighting functions, e.g. normalisation matrices or variable uniform quantisers
• CPC Class Codes: H04N 19/60 using transform coding

Abstract

In a method and apparatus for transmitting a digital image over a limited bandwidth communication channel, an image is block transformed to produce blocks of transform coefficients; the transform coefficients are quantized in accordance with a model of the visibility of quantization error in the presence of image detail; the quantized coefficients are encoded with a minimum redundancy code; and the coded, quantized transform coefficients are transmitted.

Claims

1. A transmitter for compressing and transmitting a digital image over a limited bandwidth communication channel, comprising:

a. means for performing a two-dimensional spatial frequency block transformation on the digital image to produce blocks of transform coefficients;

b. means for quantizing the transform coefficients in accordance with a model of the visibility of quantization error in the presence of image detail;

c. means for encoding the quantized transformation coefficients with a minimum redundancy code; and

d. means for transmitting the encoded transform coefficients.

2. The transmitter claimed in claim 1, wherein said means for quantizing the transform coefficients, comprises:

a. means for normalizing the coefficients in accordance with a model of the visibility of quantization errors in the presence of image detail; and

b. means for quantizing the normalized coefficients.

3. The transmitter claimed in claim 1, further including means for globally quantizing the transform coefficients based on a model of the human visual response to the spatial frequencies represented by the transform coefficients.

4. The transmitter claimed in claim 1, wherein said means for quantizing transform coefficients, comprises;

a. means for arranging the coefficients from a block into a one-dimensional array in order of decreasing spatial frequency; and

b. means for sequentially quantizing the coefficients in the array, starting with the coefficient representing the lowest frequency including;

(1) means for forming an estimate of the contrast of the image structure in the block from the previous coefficient values in the array; and

(2) means for determining the quantization for the current coefficient as a function of the contrast estimate, said function relating the contrast estimate to the visibility of quantization error in the presence of image detail having such contrast.

5. The transmitter claimed in claim 4, wherein the previous coefficients employed to form an estimate of the contrast of the image detail are the quantized coefficients, whereby the quantization values may be recovered from the quantized signal value at a receiver without error.

6. The transmitter claimed in claim 4, wherein said means for sequentially quantizing provides a predetermined quantization for the first m coefficients.

7. The transmitter claimed in claim 6 wherein said means for sequentially quantizing the coefficients includes means for detecting the presence of an edge separating uniform image areas in the block, and means for providing a predetermined quantization for all the coefficients in the array when such an edge is detected.

8. A method for compressing a digital image for transmission over a limited bandwidth communication channel, comprising the steps of:

a. performing a two-dimensional spatial frequency block transformation on the digital image to produce blocks of transform coefficients;

b. quantizing the transform coefficients in accordance with a model of the visibility of quantization error in the presence of image detail; and

c. encoding the quantized transformation coefficients employing a minimum redundancy code.

9. The method claimed in claim 8, wherein said step of quantizing the transform coefficients, comprises the steps of:

a. normalizing the coefficients in accordance with a model of the visibility of quantization errors in the presence of image detail; and

b. quantizing the normalized coefficients.

10. The method claimed in claim 8, further including globally quantizing the transform coefficients based on a model of the human visual response to the spatial frequencies represented by the transform coefficients.

11. The method claimed in claim 8, wherein said step quantizing transform coefficients, comprises the steps of:

a. arranging the coefficients from a block into a one-dimensional array in order of increasing spatial frequency; and

b. sequentially quantizing the coefficients in the array, starting with the coefficient representing the lowest frequency including the steps of;

(1) forming an estimate of the contrast of the image structure in the block from the previous coefficient values in the array; and

(2) determining the quantization for the current coefficient as a function of the contrast estimate, said function relating the contrast estimate to the visibility of quantization error in the presence of image detail having such contrast.

12. The method claimed in claim 11, wherein the previous coefficients employed to form an estimate of the contrast of the image detail are the quantized coefficients, whereby the quantization values may be recovered from the quantized signal value at a receiver without error.

13. The method claimed in claim 11, wherein said steps of sequentially quantizes the first m coefficients with a predetermined step size.

14. The method claimed in claim 13 wherein said step of sequentially quantizing the coefficients includes detecting the presence of an edge separating uniform image areas in the block, and quantizing for all the coefficients in the array with a predetermined step size when such an edge is detected.