Image analysis – Image compression or coding
Reexamination Certificate
1997-05-13
2002-07-30
Boudreau, Leo (Department: 2621)
Image analysis
Image compression or coding
C382S246000, C382S166000
Reexamination Certificate
active
06427025
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an image processing apparatus, such as an image data compressing or expanding apparatus, and an image processing method which is employed in such an apparatus.
2. Related Background Art
Regarding the images created on a computer by, for example, DTP, there is a demand for a higher image quality in recent years, and development of color or multi-level images has advanced. The quantity of data in such an image quality is about 88 M bytes in the case of an image having, for example, A4 size, 4000 dpi, and 256 “gray” levels (gradations) for each of three colors.
Handling image data as code data, such as in a page description language, can reduce the quantity of data. However, it takes time for code data to be transformed into image data. In addition, it may be impossible to reproduce original image data from the transformed code data.
Color multilevel images are generally compressed by the ADCT (adaptive discrete cosine transform) method recommended by JPEG (Joint Photographic Expent Group). This ADCT compression method will be described below in detail.
FIG. 26
is a block diagram illustrating a functional configuration of an ADCT compressing apparatus.
In the figure, reference numeral
3101
denotes a color space conversion unit for converting a color space (NTSC-RGB) by signals of respective colors (red (R), green (G) and blue (B)) of the NTSC method into a color space (YCrCb) represented by a luminance signal Y and two color difference signals Cr and Cb.
Reference numeral
3102
denotes a sub-sampling unit for reducing color difference data by utilizing the characteristics of human eyes in that they are sensitive to a luminance and insensitive to a color difference. In a practical operation, the sub-sampling unit
3102
reduces the quantity of color difference data to one half of the original quantity by obtaining the average value of adjacent two color difference data. Reference numeral
3103
denotes a DCT unit for dividing the image data input through the sub-sampling unit
3102
into blocks each having dimensions of 8×8 in the horizontal and vertical directions and for conducting DCT on each of the obtained blocks to transform the input image into a frequency space.
Reference numeral
3104
denotes a quantization unit for dividing 64 DCT coefficients by quantized values having different step widths. Reference numeral
3105
denotes a Huffman encoding unit for dividing the 64 quantized DCT coefficients into a single DC coefficient and 63 AC coefficients and for coding the respective coefficients according to the Huffman table recommended by JPEG. After a header, such as quantization table data or Huffman table data, is affixed to the coded data, the coded data is stored in a memory or transmitted to another apparatus.
FIG. 27
is a block diagram illustrating a function configuration of an ADCT expanding apparatus.
In the figure, reference numeral
3205
denotes a Huffman decoding unit for decoding input coded data to generate quantized data. Reference numeral
3204
denotes an inverse quantization unit for transforming the quantized data generated by the Huffman decoding unit
3205
into DCT coefficient data. This is achieved by multiplying 64 coefficients by the quantized values using the quantization table data employed by the quantization unit
3104
for quantization.
Reference numeral
3203
denotes an inverse DCT unit for performing inverse DCT on the DCT coefficient data obtained by the inverse quantization unit
3204
to obtain actual image data. Reference numeral
3202
denotes an interpolation unit for interpolating the data Cr and Cb which have been removed by the sub sampling unit
3102
during data compression by the simple repeating method. Reference numeral
3201
denotes a color space editing unit for converting the data YCrCb into NTSC-RGB data or color space data suitable for that particular apparatus.
The flow of the process in a conventional data compressing or expanding apparatus will now be described.
FIG. 28
illustrates part of image data of a color multilevel image created on a computer. The image data shown in
FIG. 28
is NTSC-RGB data which shows the portion (16 pixels×16 pixels) of the image data representing a character portion of a color multilevel image.
In
FIG. 28
, reference numeral
3301
denotes red (R) data, reference numeral
3302
denotes green (G) data, and reference numeral
3303
denotes blue (B) data. The data of each pixel is obtained by a 8-bit register (0 to 255). The data shown in
FIG. 28
represents the portion of an image in which a character of blue (R, G, B)=(34, 30, 225) is written on a background of a slightly dark white (R, G, B)=(225, 225, 225).
The color space conversion unit
3101
performs conversion from NTSC-RGB to YCrCb by the following equations:
Y=
0.299
×R+
0.587
×G+
0.114
×B
Cr=
0.713 (
R−Y
)
Cb=
0.564 (
B−Y
)
In addition, in order to allow for overshoot or undershoot during the operations in the YCrCb color space, the color space conversion unit
3101
performs rounding of data in conformity with the recommendation of CCIR as follows:
Y=
219.0
×Y+
16.5
Cr=
224.0
×Cr+
128.5
Cb=
224.0
×Cb+
128.5
The sub-sampling unit
3102
performs sub-sampling on the color difference data Cr and Cb in the data of YCrCb obtained by the above-described equations using the average value method as the sub-sampling method. In this method, the average value of the data of the adjacent two pixels is obtained to obtain single data.
FIG. 29
illustrates the data obtained through the color space conversion unit
3101
and the sub-sampling unit
3102
. In
FIG. 29
, reference numeral
3401
denotes Y data (luminance data), reference numeral
3402
denotes Cr data (color difference data) and reference numeral
3403
denotes Cb data (color difference data). As can be seen in
FIG. 29
, the quantity of each of Cr and Cb data has been reduced to one half by the sub-sampling unit
3102
.
Y data
3401
, Cr data
3402
and Cb data
3403
, shown in
FIG. 29
, are input to the DCT unit
3103
. The DCT unit
3103
divides the data into blocks each of which contains data having dimensions of 8×8 in the horizontal and vertical directions. As a result, Y data
3401
is divided into four blocks
3401
a
through
3401
d
. Similarly, Cr data
3402
is divided into two blocks
3402
a
and
3402
b
. DCT is performed on each of these eight blocks.
FIG. 30
illustrates the data obtained by performing DCT conversion on the eight blocks shown in FIG.
29
. In
FIG. 30
, reference numeral
3501
denotes data obtained by performing DCT on Y data
3401
. The four blocks
3401
a
through
3401
d
respectively correspond to blocks
3501
a
through
3501
d
. Similarly, reference numeral
3502
denotes data obtained by performing DCT on Cr data
3402
. Reference numeral
3503
denotes data obtained by performing DCT on Cb data
3403
. 64 coefficients of each of the blocks obtained by DCT include a single DC component (in the left upper corner) and 63 AC components.
Next, the quantization unit
3104
quantizes the DCT data
3501
through
3503
shown in FIG.
30
.
The quantization table recommended by JPEG is employed as the quantization table for quantization.
FIG. 35
illustrates that quantization table. In
FIG. 35
, reference numeral
4001
denotes data for Y components, and reference numeral
4002
denotes data for Cr and Cb components.
FIG. 31
illustrates data obtained by quantization. Reference numeral
3601
denotes data obtained by quantizing Y data. Reference numeral
3602
denotes data obtained by quantizing Cr data. Reference numeral
3603
denotes data obtained by quantizing Cb data.
The Huffman coding unit
3105
divides each of the quantized data
3601
through
3603
into a DC component and AC components. Regarding the DC component, the Huffman coding unit
3103
creates the optimum Huffman coding table by obtaining
Saito Kazuhiro
Shimomura Yukari
Boudreau Leo
Canon Kabushiki Kaisha
Dang Duy M.
Fitzpatrick ,Cella, Harper & Scinto
LandOfFree
Image processing with selection between reversible and... 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 processing with selection between reversible and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image processing with selection between reversible and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2913362