Image analysis – Image compression or coding – Predictive coding
Reexamination Certificate
2001-02-28
2004-12-14
Boudreau, Leo (Department: 2621)
Image analysis
Image compression or coding
Predictive coding
C382S236000
Reexamination Certificate
active
06832004
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an image coding apparatus and an image coding method for compressively coding image data.
2. Description of the Related Art
Image data generally falls into two broad categories: artificial image data generated by computer such as image data described in PDL (page description language), and scan-in image data read by scanner such as bit map image data. These two kinds of image data are most often coded compressively in order to reduce their quantities.
Illustratively, JPEG-LS (Part-1 ISO/IEC FDIS 14495-1) is known as a coding technique applicable to both artificial image data and scan-in image data. JPEG-LS has been standardized as a method for improving by about 10 percent the coding efficiency achieved by JPEG (Joint Photographic Expert Group) and JBIG (Joint Bi-level Image Experts Group) schemes. A major characteristic of the JPEG-LS method is that it applies lossless (reversible) coding to artificial image data and near-lossless (near-reversible) coding to scan-in image data.
FIG. 13
is a schematic block diagram of a typical image coding apparatus adopting JPEG-LS for image data coding. This image coding apparatus works as follows: input image data
401
and decoded pixel data
402
are first fed to a window composing element
41
. In turn, the window composing element
41
sends a group of peripheral pixel values
403
to a pixel value gradient determining element
42
. Given the peripheral pixel values, the pixel value gradient determining element
42
computes a pixel value gradient of the peripheral pixels surrounding a target pixel to be coded. If the computed gradient is found less than a tolerance
404
coming from a tolerance designating element
43
, the target pixel to be coded is judged flat and control is passed on to a run length extracting element
44
. If the gradient is found greater than the tolerance
404
, then regular mode is entered. That is, a pixel value predicting element
45
predicts a value of the target pixel to be coded on the basis of the peripheral pixel value group
403
. A prediction error computing element
46
computes a prediction error regarding the target pixel. A prediction error quantizing element
47
quantizes the computed prediction error. Taking the quantized prediction error as a coding symbol
406
, an information source coding element
48
codes the prediction error. The coded output of the element
48
combines with a code from the run length extracting element
44
to become output coded data
407
. A local decoding element
49
performs local decoding and supplies the result as decoded pixel data
402
to the window composing element
41
.
The JPEG-LS method outlined above is based on pixel value prediction and has no recourse to resolution adaptation known as sub-sampling. Sub-sampling involves discarding the whole information furnished by any one pixel. In low frequency image regions, particularly in graduated portions of scan-in image data, sub-sampling should preferably be utilized to enhance the efficiency of coding. In the setup of
FIG. 13
, however, the pixel value gradient determining element
42
determines whether or not to proceed with sub-sampling based on the gradient of pixels around the target pixel to be coded. In other words, the sub-sampling of the setup has no direct relation to prediction error.
Meanwhile, some of the image data coding techniques proposed so far have introduced the concept of sub-sampling into predicted coding. Illustratively, Japanese Published Unexamined Patent Application No. Hei 10-294872 discloses a sub-sampling-based coding technique for selecting an index (identification information) of one of multiple predictors that has the smallest prediction error so that a run length of the rank of the selected index is coded.
More particularly, as shown in
FIG. 14
, a group of peripheral pixel values
503
representing pixels around a target pixel to be coded is acquired through a window composing element
51
. Given the peripheral pixel value group
503
, a pixel value approximating element
52
having multiple predictors computes a group of decoded pixel value candidates
504
. An error selecting element
54
checks for a difference between each of the pixel values approximated by the pixel value approximating element
52
on the one hand, and the value of a target pixel in an image of interest on the other hand. If the difference (i.e., prediction error) is found less than a tolerance
505
coming from a tolerance designating element
53
, then the error selecting element
54
sends an index of the predictor having the smallest prediction error to a decoded pixel selecting element
55
and a rank/run length extracting element
56
. If the prediction error is judged to be greater than the tolerance
505
, the error selecting element
54
supplies the rank/run length extracting element
56
with an index of an error minimizing approximation
506
for minimizing the prediction error as well as with the value of the prediction error in effect at that point. Based on the error minimizing approximation index
506
or the prediction error value received from the error selecting element
54
, the rank/run length extracting element
56
extracts a run length of the index rank and sends it as a coding symbol
507
to an information source coding element
57
. The information source coding element
57
performs information source coding to compress the statistical redundancy included in the coding system
507
to provide output coded data
508
. The decoded pixel selecting element
55
, given the error minimizing approximation index
506
, selects a decoded pixel value from the group of decoded pixel value candidates
504
and feeds it as decoded pixel data
502
to the window composing element
51
for decoding of the next pixel.
In the manner described above, the coding technique of Japanese Published Unexamined Patent Application No. Hei 10-294872 adopting the concept of sub-sampling subjects artificial image data and scan-in image data to the same framework of reversible and near-reversible coding, and executes an image coding process by coding indexes of prediction techniques (predictors) for approximating each target pixel to be coded. That is, the approximation techniques (predictors) based on important factors of images are adopted for data compression without deterioration of image quality.
However, the conventional image data coding method disclosed by the above-cited patent application has its share of deficiencies. The method, in selecting one of the prediction techniques (predictors) for approximating the target pixel to be coded, abides by a single rule dictating that the selected technique be one that has the smallest prediction error. Hence the possible major flaw of the method as outlined below.
With pixel values making up artificial image data, efforts to reduce their prediction errors translate into improvements in image quality of non-reversible coding. Pixel values constituting scan-in image data, on the other hand, undergo MTF (modulation transfer function) illustratively in the optics of a scanner and otherwise gather various noises along the way. For that reason, minimizing the prediction error of each pixel does not necessarily result in improvements in image quality of non-reversible coding.
The above-cited conventional image data coding technique, particularly when addressing scan-in image data and when their prediction errors exceed a tolerance, is liable not only to worsen its coding efficiency but also to slow down its processing speed. That is because the prediction errors themselves are coded. Even if the prediction errors are below the tolerance, the selected predictor tends to change frequently as long as the sole judging rule dictates that the prediction error be minimal per pixel. The tendency defeats the supposed effects of coding run lengths of index ranks, rendering the coding process inefficient. Furthermore, improvements in image quality cannot be expected unless predict
Koshi Yutaka
Sekino Masanori
Shishido Shinji
So Ikken
Boudreau Leo
Dang Duy M.
Fuji 'Xerox Co., Ltd.
Morgan & Lewis & Bockius, LLP
LandOfFree
Image coding apparatus and image coding method does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image coding apparatus and image coding method, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image coding apparatus and image coding method will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3334524