Image analysis – Image compression or coding – Parallel coding architecture
Reexamination Certificate
1999-02-09
2003-04-15
Patel, Jayanti K. (Department: 2621)
Image analysis
Image compression or coding
Parallel coding architecture
C348S714000, C382S246000, C382S305000
Reexamination Certificate
active
06549667
ABSTRACT:
BACKGROUND OF THE INVENTION
This invention relates to an image data coding apparatus for compressing image data, an image data reconstructing method and an apparatus for reconstructing images from compressed data. More particularly, it relates to a coding apparatus for coding continuous tone images, after dividing them into blocks each comprising a plurality of picture elements, by orthogonally transforming the picture elements in respective blocks, and to an image data reconstructing method and apparatus for reconstructing images from the orthogonally transformed coded data.
Gradient values for respective picture elements need to be coded highly efficiently, for storing or transmitting, at high speed and high quality, image data, especially grayscale image data and color image data, whose information volume is exceedingly greater than that of coded numerical data.
DESCRIPTION OF THE RELATED ART
Conventionally, an adaptive discrete cosine transform coding method is used as a highly efficient method for compressing image data, for example.
The following is a description of the adaptive discrete cosine transform coding method(, or the ADCT method in short).
According to the ADCT method, images are divided into blocks comprising e.g. 8×8 picture elements. DCT coefficients expressing the distribution of space frequencies are determined by performing two-dimensional discrete cosine transform (hereafter abbreviated as DCT) for the image signals from respective divided blocks. The image signals are coded by quantizing the DCT coefficients using thresholds corresponding to visions, according to a Huffman table obtained statistically from the quantized coefficients.
FIG. 1
is a block diagram of a coding circuit per the ADCT method.
FIG. 2
shows exemplary data for an original image received by the two-dimensional DCT unit.
FIG. 3
shows exemplary DCT coefficients representing a space frequency distribution.
The coding operation per the ADCT method is described below.
A two-dimensional DCT unit
24
receives from an input terminal
23
sixty-four (64) image signals, such as those shown in
FIG. 2
, for a black comprising 8×8 picture elements. By orthogonally transforming the received image signals, the two-dimensional DCT unit
24
transforms them into coefficients having space frequency distributions such as those shown in
FIG. 3
, thereby calculating DCT coefficients, and outputs them to a linear quantizer
25
.
FIG. 4
is a block diagram of the two-dimensional DCT unit
24
. A one-dimensional DCT unit
30
performs one-dimensional DCTs for the image signals inputted from the input terminal
23
, and a transposer
31
transposes the matrix obtained from the one-dimensional DCT unit
30
. Then, a one-dimensional DCT unit
32
performs one dimensional DCTs, as with the one-dimensional DCT unit
30
described earlier. As with the transposer
31
described earlier, a transposer
33
transposes the matrix obtained at the one-dimensional DCT unit
32
and outputs it to a terminal
34
. Through similar performances for all of image blocks, the inputted image signals are transformed to the DCT coefficients.
FIG. 5
is a table of thresholds for DCT coefficients in a block.
FIG. 6
is a table of exemplary quantized DCT coefficients.
Further explanation is made by referring back to FIG.
1
. The linear quantizer
25
quantizes the inputted DCT coefficients by the quantization matrix comprising the optically determined thresholds shown in FIG.
5
. The linear quantization produces quantized coefficients such as those shown in
FIG. 6
, which indicate that the DCT coefficients smaller than thresholds become 0, thereby forming quantized coefficients in which the quantized DC elements and only a few quantized AC coefficients have non-zero values.
Generally, the higher the visual sensitivities the lower the space frequencies are, and the lower the visual sensitivities the higher the space frequencies are. Therefore, as shown in
FIG. 5
, the absolute values of the quantization matrix elements, i.e. the thresholds for DCT coefficients corresponding to lower space frequencies, are smaller, and the absolute values corresponding to higher space frequencies are larger. Accordingly, as shown in
FIG. 6
, of quantized coefficients, only the quantized DC element at the top left end and a very few quantized AC coefficients representing low space frequency elements become significant coefficients having non-zero value values, whereas all others become insignificant coefficients, in most cases.
FIG. 7
shows the order of scanning the generated quantized coefficients (from the lower frequency elements to the higher frequency elements of the space frequencies).
The linear quantizer
25
transforms the two-dimensionally arrayed quantized coefficients into a one-dimensional form according to the zigzag scanning order shown in FIG.
7
and inputs them to a variable length coder
26
in the next stage.
Upon receipt of these quantized coefficients, the variable length coder
26
codes generated quantized coefficients by referring to a coding table
27
composed of a Huffman table created by statistical volumes for the respective images. As for their quantized DC elements, the differences between the current quantized DC coefficients and the previous quantized DC coefficients are coded in variable lengths. As for their quantized AC coefficients, the values of the quantized AC coefficients (hereafter referred to as indices) of significant coefficients (non-zero value coefficients) and the run lengths (hereafter referred to as runs) of the insignificant coefficients (zero value coefficients) are coded in variable lengths. The output terminal
28
sequentially outputs the coded data to the external units.
Meanwhile, the coded data obtained by the coding circuit per the ADCT method are reconstructed as images according to the following method.
FIG. 8
is a block diagram of a decoding circuit per ADCT method.
A variable length decoder
41
receives the coded data inputted from an input terminal
40
. The variable length decoder
41
decodes the received coded data into fixed length data of the indices and runs and outputs the decoded data to a dequantizer
43
, according to a decoding table
42
formed by a table inverse of the Huffman table composing the coding table
27
.
On receiving the decoded data (the decoded quantized coefficients), the dequantizer
43
reconstructs the dequantized DCT coefficients through a dequantization by multiplying the respective decoded data by the thresholds stored at the corresponding positions in a quantization matrix
48
. The dequantizer
43
outputs the dequantized DCT coefficients to a two-dimensional inverse DCT unit
44
.
The two-dimensional inverse DCT unit
44
orthogonally transforms the received dequantized DCT coefficients indicating the distribution of the space frequencies into image signals.
The two-dimensional inverse DCT unit
44
is explained in further detail.
FIG. 9
is a block diagram of a two-dimensional inverse DCT unit of the ADCT decoding circuit.
A one-dimensional inverse DCT unit
51
performs one-dimensional inverse DCTs for the DCT coefficients inputted from a terminal
50
and outputs them to a transposer
52
. The transposer
52
transposes the matrix representing the outputs from the one-dimensional inverse DCT unit
51
. A one-dimensional inverse DCT unit
53
again performs one-dimensional inverse DCTs on the transposed matrix obtained from the matrix transposition at the transposer
52
. As with the transposer
52
, a transposer
54
transposes the matrix representing the outputs from the one-dimensional inverse DCT unit
53
. A terminal
45
outputs signals obtained by these processes, thereby reconstructing the images.
Per the earlier described ADCT method, the quantized coefficients are obtained by quantizing DCT coefficients by quantization thresholds.
FIG. 10
is a block diagram of a conventional linear quantization circuit.
DCT coefficients inputted from a terminal
60
are supplied to a DCT coefficient receptor
64
to be stored. Th
Chiba Hirotaka
Fukuda Masahiro
Murashita Kimitaka
Noda Tsugio
LandOfFree
Image data encoding method and 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 data encoding method and apparatus, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image data encoding method and apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3087819