Image analysis – Image enhancement or restoration – Edge or contour enhancement
Reexamination Certificate
2000-04-17
2004-05-25
Johnson, Timothy M. (Department: 2623)
Image analysis
Image enhancement or restoration
Edge or contour enhancement
C382S197000, C382S199000, C382S250000
Reexamination Certificate
active
06741752
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a method of compressing and decompressing image signals and, more particularly, to a method for restoring block-coded images much closer to their original images.
2. Description of the Related Art
Coding techniques based on Block Discrete Cosine Transform (BDCT) or Vector Quantization (VQ) have been successfully applied to compression of still and moving images.
In BDCT- or VQ-based coding, such as JPEC (Joining Photographic Expert Group) method, an image is segmented into plural non-overlapping blocks each of which contains N pixels horizontally by N pixels vertically (i.e. N×N), as shown in
FIG. 1
, where N is usually
8
.
In BDCT encoding process, original image information of the respective segmented blocks is transformed and compressed into DCT coefficients in accordance with DCT algorithm. The DCT coefficients are quantized, and are encoded in a form suitable for transmission via a channel or storage in a medium. In BDCT decoding process, data received or read from a channel or medium is decoded and dequantized, and undergoes inverse DCT so as to restore its original image.
However, the blockwise operation of the coding algorithm causes undesirable blocky artifacts, which may be visible grey-level discontinuities along block boundaries (refer to
FIG. 8
a
). The blocky artifacts (or interblock grey-level discontinuities) result from visually obvious level difference between blocks when image signals are compressed with DCT or VQ because, during the compression, noises in boundary blocks are greater than those in non-boundary blocks.
Conventional postprocessing techniques to reduce such artifacts are based on the known facts that “blocky noise” (this term refers to ‘quantization noise’ causing blocky artifacts) is caused by amplified differences between boundary pixel values of neighboring blocks and contains undesirable high frequency components. In other words, most of conventional blocky-artifact-reduction techniques can be classified into two categories: (1) adjustment of block boundary pixels to reduce interblock discontinuities, and (2) lowpass filtering to reduce the high frequency components of the blocky noise.
The first approach is described in, for example, an article by Yang et al., entitled “Regulated reconstruction to reduce blocking effects of block discrete cosine transform compressed images” (IEEE Trans. on Circuits and Systems for Video Technology, vol. 3, pp. 421-432. December 1993). Other examples are founded in articles “Nonlinear constrained least squares estimation to reduce artifacts in block transform-coded images” (M. Crouse et al.,
Proc. International Conference on Image Processing
, pp. 462-465, 1995) and “Blocky artifact reduction using an adaptive constrained least squares method” (J. Yang et al.,
Electronic Letters
, pp. 854-855, vol. 33, no. 10, May 1997). These methods reduce intensity discontinuity across block boundaries, but they often leave visible discontinuities among non-boundary pixels of blocks.
The second approach are disclosed in, for example, articles “Nonlinear space-variant postprocessing of block coded images” (B. Ramamurthi et al, IEEE Trans.
on Acoustics, Speech, and Signal Processing
, vol. 34, pp. 1258-1267, October 1986), “Iterative procedures for reduction of blocking effects in transform image coding” (R. Rosenholts et al., IEEE Trans.
on Circuits and Systems for Video Technology
, vol. 2, pp. 91-95, March 1992), “Comments on Iterative procedures for reduction of blocking effects in transform image coding” (S. J. Reeves et al., IEEE Trans.
on Circuits and Systems for Video Technology
, vol. 3, pp. 439-440. December 1993), and “A POCS(Projection On Convex Set)-based post-processing technique to reduce blocking artifacts in transform coded images” (H. Paek et al, IEEE Trans.
Image Processing
, vol. 8, no. 3, June 1998). These methods achieve smoothing in a frequency domain by attenuating a specific range of high-frequency components in a transform of a given image, but they tend to blur edges in the original images.
SUMMARY OF THE INVENTION
Accordingly, the present invention is provided to mitigate both the shortcomings stated above.
It is an object of the present invention to provide a method of eliminating interblock noise components in block coded images without deteriorating clearness of the images.
It is another object of the present invention to provide a method of effectively eliminating block noise components in block coded images without discontinuity among non-boundary pixels of blocks.
It is yet another object of the present invention to provide a block noise elimination method to yield clear edges of decoded images.
The foregoing and other objects are achieved as is now described.
A block noise elimination method of the invention uses one-dimensional pixel vectors made of pixel rows or columns across two neighboring blocks to analyze blocky artifacts as noise components residing across the neighboring blocks. Blocky noise in each pixel vector is modeled as shape vector multiplied by boundary discontinuity. The shape vector, varying with local image activities, is estimated by Minimum Mean Squared Error (MMSE) approach. That is, MMSE-based noise estimation is adopted to utilize statistical characteristics of noise components in DCT coded images. For this purpose, blocky artifacts are analyzed as noise components in pixel vectors made of pixel rows and/or columns across the neighboring blocks. The shape vectors, depending upon the local image activities, are estimated by MMSE approach from the statistics of the noise vectors prior to postprocessing. Then, in accordance with the shape vector estimation algorithm of the present invention, boundary discontinuity and local image activity parameter for each pixel vector is calculated and corresponding blocky noise is eliminated.
In accordance with a preferred aspect of the present invention, to eliminate block boundary noise components residing across two neighboring blocks in a block-encoded image such as a BDCT-coded or a VQ-coded image, the block boundary noise components are estimated statistically from encoded pixel vectors of plural various images in an encoding part. The block boundary noise components are estimated using block boundary discontinuity parameters of the pixel vectors and local image activity parameters of the pixel vectors. The block boundary discontinuity parameters are defined as slopes of pixel values across block boundary. The local image activity parameters are defined as nonzero coefficient indices of highest frequency in N-point DCT coefficient vectors of the respective pixel vectors. The estimated noise components are transferred to a decoding part. The estimated block boundary noise components may be transferred to a decoding part together with the encoded pixel vectors in real time. Alternatively, the estimated block boundary noise components are previously stored in a table within a decoding part. The block boundary noise components are denoted by shape vectors representing shapes of the block boundary noises across the adjacent blocks. The shape vectors are one-dimensional vectors corresponding to pixel rows or columns of the adjacent blocks. The shape vectors are estimated with respect to local image activity parameters. The shape vectors are estimated using original pixel vectors of an encode image, encoded pixel vectors and block boundary discontinuity parameters of the encoded pixel vectors.
REFERENCES:
patent: 5337088 (1994-08-01), Honjo
patent: 5590064 (1996-12-01), Astle
patent: 5796875 (1998-08-01), Read
patent: 5960388 (1999-09-01), Nishiguchi et al.
patent: 5978514 (1999-11-01), Yamaguchi et al.
patent: 5982442 (1999-11-01), Guntzburger
patent: 6064954 (2000-05-01), Cohen et al.
patent: 6463100 (2002-10-01), Cho et al.
patent: 6510407 (2003-01-01), Wang
patent: 6614938 (2003-09-01), Yamaguchi et al.
Choobin Barry
F. Chau & Associates LLC
Johnson Timothy M.
LandOfFree
Method of removing block boundary noise components in... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method of removing block boundary noise components in..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method of removing block boundary noise components in... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3208228