Pulse or digital communications – Bandwidth reduction or expansion – Television or motion video signal
Reexamination Certificate
1999-08-27
2003-06-10
Kelley, Chris (Department: 2613)
Pulse or digital communications
Bandwidth reduction or expansion
Television or motion video signal
C375S240030
Reexamination Certificate
active
06577681
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an image coding apparatus which converts an image into frequency areas and codes the image by adaptively performing quantization.
2. Description of the Related Art
More gradations and higher resolution are required to obtain higher quality digital images. The capacity of images is represented as the product of the number of pixels and the number of gradation bits, which represents an enormous amount of information. Therefore, an image size is compressed to reduce costs of storage or transmission of images.
A variety of image coding systems are proposed. A typical one is JPEG (Joint Photographic Experts Group) Baseline system described in Pages 18 to 23 of “International Standards of Multimedia Coding” edited by Yasuda, published by Maruzen. This system will be explained using FIG.
12
.
In
FIG. 12
, the reference numerals
1
,
2
,
3
,
4
,
5
,
6
,
7
,
8
,
9
,
10
,
11
,
12
,
13
,
14
, and
15
designate an input image, a blocking circuit, a DCT (discrete cosine transform) circuit, a quantizer, a quantization table, a scan converting circuit, a significant coefficient detecting circuit, a grouping circuit, a run length counter, a two-dimensional Huffman coding circuit, a DC difference calculating circuit, a grouping circuit, a one-dimensional Huffman coding circuit, a multiplexing circuit, and an output code, respectively.
In
FIG. 12
, an input image
1
is split into blocks (hereinafter called pixel blocks) of 8×8 pixels each by the blocking circuit
2
. The pixel blocks are subjected to a DCT operation by the DCT circuit
3
and transform coefficients output as a result of the DCT operation are quantized by the quantizer
4
according to quantization step information stored in the quantization table
5
. The quantized transform coefficients can be represented as an 8-by-8 matrix. Usually, the transform coefficients are arranged in the array so that the vertical coefficients correspond to higher-order DCT coefficients as they go downward and the horizontal coefficients correspond to higher-order DCT coefficients as they go toward the right. The leftmost and topmost of the 64 transform coefficients is called a DC component or DC coefficient because it corresponds to a DC frequency area of a DCT area. The other
63
are called AC components or AC coefficients because they correspond to AC frequency areas.
The DC difference calculating circuit
11
calculates a difference between the DC coefficient and a DC component of a preceding image block and sends it to the grouping circuit
12
. The grouping circuit
12
calculates, from the DC difference value, a group number and additional bits shown in FIG.
15
. The additional bits are used to identify a DC difference value within an identical group. The number of bits of the additional bits is shown in FIG.
15
. The group number calculated in the grouping circuit
12
is converted into a Huffman code by the one-dimensional Huffman coding circuit
13
. The additional bits are sent to the multiplexing circuit
14
.
The AC coefficients quantized by the quantizer
4
are scan-converted in a zigzag scan order shown in FIG.
13
and are sent to the significant coefficient detecting circuit
7
. The significant coefficient detecting circuit
7
judges whether the quantized AC coefficients are “0” or not, and when “0”, supplies a count-up signal to the run length counter
9
to increment the counter value by one. When the AC coefficients are nonzero significant coefficients, the significant coefficient detecting circuit
7
supplies a reset signal to the run length counter
9
to reset the counter value and sends the AC coefficients to the grouping circuit
8
.
The run length counter
9
counts the length of a run of “0”s. The number NNNN of “0”s between two significant coefficients is sent to the two-dimensional Huffman coding circuit
10
. The grouping circuit
8
splits the AC coefficients into group numbers SSSS and additional bits shown in
FIG. 14
, and sends the group numbers to the two-dimensional Huffman coding circuit
10
and the additional bits to the multiplexing circuit
14
. The additional bits are used to identify a DC difference value within an identical group. The number of bits of the additional bits is shown in FIG.
14
.
The two-dimensional Huffman coding circuit
10
converts a combination of a run length NNNN and a group number SSSS into a Huffman code and sends it to the multiplexing circuit
14
.
The multiplexing circuit
14
multiplexes a DC coefficient Huffman code of one pixel block, an AC coefficient Huffman code, DC coefficient additional bits, and AC coefficient additional bits and outputs code data
15
.
This terminates a description of the JPEG Baseline coding system.
Furthermore, by changing the contents of the quantization table
5
of the standard image coding system described above, the following effects can be obtained:
(1) The amount of code of an input image can be controlled to a desired level.
(2) Quantization resistant to image degradation can be performed in accordance with the nature of an input image.
The above described two points will be described in more detail.
The first point will be described. When a quantization step width is small, the amount of code increases because probably the number of significant coefficients will increase and the absolute values of the significant coefficients will become large. Conversely, when a quantization step value is large, the amount of code decreases because probably a run length will increase due to a small number of significant coefficients and the absolute values of the significant coefficients will become small. By providing a basic quantization table and multiplying a quantization step value within the basic quantization table by a constant called a scaling factor, the value of quantization step width can be increased or decreased. When the amount of code is to be increased to enhance image quality, the scaling factor should be reduced. When the amount of code is to be reduced, the scaling factor should be set to a large value.
Generally, a larger amount of code makes image quality better and a smaller amount of code makes image quality worse. However, the limit of storage capacity and transmission capacity may limit the amount of code. In this case, the best image quality can be obtained by controlling the amount of code to a limit.
Numerous methods of calculating a scaling factor with the objective of controlling the amount of code as described above are proposed in Japanese Published Unexamined Patent Application Nos. Hei 7-107296 and Hei 7-212757, and the like.
These methods are used to obtain a scaling factor of an entire input image. However, when an image size is large, there are cases where it is necessary to split the image into smaller partial images and control the amount of code for each of the partial images. In this case, if the split partial image is viewed as one image, these methods are also applicable to the control of code amount for each of the partial images.
Next, the second point will be described. Usually, halftone images produced from photographs input through an image scanner or the like tend to be power-intensive in low-frequency portions of blocks subjected to a DCT operation. Accordingly, in order to reduce an error power after quantization, the quantization table
5
is organized to have small step widths for low frequencies of DCT coefficients and large step widths for high frequencies.
However, in the edge portion (a point where a sudden change of density values or brightness values is found) of an image, power is distributed also in high-frequency portions of blocks subjected to a DCT operation. As is already known, quantizing an edge portion with a quantization table for halftone images causes image distortion called mosquito noise. To suppress the occurrence of the image distortion, a quantization table suitable for each of image parts should be selected for quantization. Alternatively, without changing a quanti
Fuji 'Xerox Co., Ltd.
Kelley Chris
Morgan & Lewis & Bockius, LLP
Parsons Charles
LandOfFree
Image coding apparatus 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, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image coding apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3115838