Apparatus for reducing quantization noise in image signals

• US 5,636,295 A
• Filed: 03/27/1995
• Issued: 06/03/1997
• Est. Priority Date: 10/07/1992
• Status: Expired due to Fees

• Current Assignee: Daewoo Electronics Company Limited (Dayou Holdings Co.,Ltd.)
• Original Assignee: Daewoo Electronics Company Limited (Dayou Holdings Co.,Ltd.)
• Inventors: Kim, Jong-Il
• Primary Examiner(s): Couso, Jose L.
• Assistant Examiner(s): Bella, Matthew C.
• Application Number:
  US08/409,839
• Time in Patent Office: 799 Days
• Field of Search: 382/233, 382/251, 382/275, 348/405, 381/94, 364/574, 341/94, 341/143, 375/254, 358/463
• US Class Current: 382/251
• CPC Class Codes: H04N 19/86 involving reduction of coding artifacts, e.g. of blockiness
• CPC Class Codes: H04N 5/21 Circuitry for suppressing or minimising disturbance, e.g. moiré or halo

Abstract

An apparatus for decoding a discrete cosine transform coded image signal includes an image restoration circuit which comprises a first conversion block for converting the reconstructed image signal into its frequency domain image signal including image signal phase components and image signal amplitude components thereof; a predictive quantization noise generation block for responding to the quantization parameter and for generating a predictive quantization noise having a known probability density function; a second conversion block for converting the predictive quantization noise into its frequency domain noise including noise phase components and noise amplitude components; an adder circuit for generating compensated amplitude components representative of the difference between the image signal amplitude components and the noise amplitude components; and a inverse conversion block responsive to the image signal phase components and the compensated amplitude components for converting them to restore the image signal, free of the quantization noise, in the time domain to thereby improve the quality of the displayed image.

Claims

1. An apparatus for decoding a discrete cosine transform coded image signal, which includes means for inversely quantizing the discrete cosine transform coded image signal to generate discrete cosine transform coefficients by using a quantization parameter representative of a quantization step size used in a quantization process of the discrete cosine transform coded image signal, means for converting the discrete cosine transform coded coefficients into a reconstructed image signal, said reconstructed image signal having a frequency domain quantization noise, and a frame memory for storing the reconstructed image signal to be displayed, characterized in that said apparatus further includes an image restoration circuit comprising:

first conversion means for converting the reconstructed image signal into its frequency domain image signal including image signal phase components and image signal amplitude components thereof;

predictive quantization noise generation means for responding to the quantization parameter and for generating a predictive quantization noise having a known probability density function;

second conversion means for converting the predictive quantization noise into its frequency domain noise including noise phase components and noise amplitude components;

means responsive to the image signal amplitude components and the noise amplitude components for generating compensated amplitude components representative of the difference between the image signal amplitude components and the noise amplitude components, and

inverse conversion means responsive to the image signal phase components and the compensated amplitude components for converting them to restore the image signal, free of the quantization noise, in the time domain to thereby improve the quality of the displayed image.

2. The apparatus as recited in claim 1, wherein the predictive quantization noise generation means includes error modeling means responsive to the quantization parameter for converting the quantization noise into a set of Discrete Cosine Transform coefficients having a known probability density function;

means responsive to the quantization parameter for controlling the variance and the maximum amplitude level of the probability density function; and

inverse conversion means for converting the set of Discrete Cosine Transform coefficients into the predictive quantization noise.

3. The apparatus as recited in claim 2, wherein the image restoration circuit further comprising windowing means for dividing the image signal into a plurality of subimages for subimage-by-subimage processing, and inverse windowing means for combining and restoring the processed subimages from the second conversion means into the whole image signal, and wherein the predictive quantization noise generation means further includes means for dividing the predictive quantization noise into a multiplicity of predictive quantization subnoises, each having a size identical to that of the subimage signal.

4. The apparatus as recited in claim 3, wherein the variance is determined as the quantization parameter divided by 10.