JP3382358B2 - Image processing apparatus and method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus and method, and more particularly to raster block conversion of image data.

[0002]

2. Description of the Related Art Conventionally, as a compression technique for multi-valued images,
International standard JPEG (Joint
Photographic Expert Group) method is available. This compression method uses YUV for RGB pixel components input in raster order.
Pixel component (Y indicates luminance, U and V indicate chromaticity)
The resolution is reduced by subsampling in some cases.

This YUV image data is subjected to discrete cosine transform (DCT transform) in 8 × 8 block units to become spatial frequency components. Those converted into spatial frequencies are called DCT coefficients. Then, the DCT coefficient is calculated by two types of 8 × 8 quantization tables of a luminance component (Y) and a chromaticity component (U, V).
Each is quantized in 8 × 8 units. The quantized coefficient here is based on the Huffman coding method which is a variable length coding method.

As described above, in the DCT conversion, processing is performed in 8 × 8 block units. Therefore, pixels input in raster order must be converted into block sequential processing before the DCT conversion processing. In order to realize this processing by hardware, a memory for at least eight lines is prepared, and this conversion is realized by changing the reading order for the pixel data once written. Further, in order to perform this conversion processing in a pipeline manner, it is general that two 8-line memories are prepared, and writing and reading are simultaneously performed.

[0005]

However, in the conventional conversion processing described above, in order to realize the raster block conversion in a hardware pipeline, two memories for eight lines are prepared as described above. However, there is a problem in that this method leads to an increase in the cost of the device.

Specifically, in order to perform compression / expansion of image data in a pipeline manner by hardware, writing and reading to / from a line buffer for raster block conversion must be executed at the same time. It is necessary to prepare two buffers. In other words, when writing to the first line buffer, reading is performed from the second line buffer, and when raster block conversion for 8 lines is completed, this time, conversely, from the first line buffer, The operation of reading and writing to the second line buffer must be repeated.

The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to perform raster block conversion of image data using a single line buffer.

[0008]

The present invention has the following structure as one means for achieving the above object. The image processing apparatus according to the present invention uses a single line buffer having a storage size for a predetermined number of lines (n lines) and reading / writing of image data in the line buffer from the storage size for one line in the line buffer. And a conversion means for performing raster block conversion on the image data by controlling a data group of a predetermined size (m pixels) which is smaller than the above, and the conversion means inputs the raster order or the block order of m × n pixels. The image data to be stored is written in the line buffer for the storage size of the line buffer, in the data group unit, and the image data held in the line buffer is in the data group unit, in a block order of m × n pixels. Or, read in raster order, and immediately after that,
Reading and writing of the line buffer for writing the next input image data to the read address in the unit of the data group is repeated until the input image data is exhausted, and the line buffer is stored for the storage size of the line buffer. Of the image data is read in the data group unit so as to be in a block order of m × n pixels or a raster order.

According to the image processing method of the present invention, the read / write of image data of a single line buffer having a storage size for a predetermined number of lines (n lines) is smaller than the storage size for one line of the line buffer. Predetermined size (m
Image data input in raster order or m × n pixel block order is stored in the line buffer. For each size, the data group is written in the line buffer, and the image data held in the line buffer is read in the data group unit in a block order of m × n pixels or a raster order, and the reading is performed. Immediately after that, reading and writing of the line buffer for writing the next input image data to the read address in the unit of the data group is repeated until the input image data is exhausted, and the storage of the line buffer held in the line buffer. The image data for the size is divided into the data group unit so that it is arranged in a block order of m × n pixels or a raster order. Characterized in that the read to.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings. [First Embodiment] FIG. 1 is a block diagram showing the arrangement of a raster block converter in an image processing apparatus (hereinafter referred to as apparatus) according to the first embodiment of the present invention. Here, an address generation method in this raster block conversion unit will be described with reference to FIG.

In FIG. 1, reference numeral 1 is an address generator for raster block conversion according to the present apparatus, and 2 is an address generator.
Line buffer section for raster block conversion, 3
Is a color conversion unit for converting pixels of RGB components into YUV components. Further, 4 is a DCT unit for performing a discrete cosine transform in units of 8 × 8 pixels, 5 is a quantizer for quantizing the DCT coefficient DCT-transformed by the DCT unit 4, and 6
Is a coding unit for coding the quantized coefficient.

In the raster block conversion section shown in FIG. 1, the input image data is, as shown in FIG. 2, starting from the upper left portion of the image 20, moving from the left direction to the right direction, and further from the upper direction. It is input in raster order downward. Since the image data input in raster order in this way is DCT-converted by the DCT unit 4 in units of 8 × 8, as shown in FIG. From the right to the top and from the top to the bottom, and then into the block sequence of the second block, the third block, the fourth block, ... here,
The raster block conversion is achieved by temporarily storing the raster-sequentially input image data in the 8-line line buffer unit 2 and reading out the read-out sequence like the first 8 lines shown in FIG.

The image data, which has been converted into blocks sequentially through the line buffer unit 2, is sequentially sent to the color conversion unit 3, where RGB → YUV conversion is performed. further,
The DCT unit 4 performs discrete cosine transform, the quantization unit 5 quantizes,
Encoding is performed by the encoding unit 6, respectively, and as a result,
Compressed data is generated. Further, FIG. 4 is a diagram for explaining compressed data decompression in the apparatus according to the present embodiment. In the figure, reference numeral 11 is the encoding unit 6 shown in FIG.
Is a decoding unit that performs the inverse operation of the above, and similarly, the inverse quantization unit 12 for the quantization unit 5 and the inverse DC for the DCT unit 4
The inverse color conversion unit 14 exists as a component with respect to the T unit 13 and the color conversion unit 3. Then, the compressed data obtained by the device shown in FIG. 1 is decoded in the path completely opposite to the case of compression, that is, decoded by the decoding unit 11 shown in FIG. The inverse DCT unit 13 performs the inverse discrete cosine transform, and the inverse color transform unit 14 performs the inverse color transform. The pixels converted into raw image data in this way are converted from block-sequential to raster-sequential via the line buffer unit 15.

In the apparatus according to the present embodiment, since one line buffer realizes raster block conversion, reading and writing are performed in one processing cycle. That is, the same address is read at a certain clock and the write operation is performed half a clock later. The apparatus according to the present embodiment is involved in address generation in the processing at this time, and performs pipeline conversion with one line buffer in a pipeline by generating an address according to a processing routine described later.

Raster address conversion in the apparatus according to this embodiment will be described below. FIG. 5 is a flowchart showing a raster address conversion procedure in the device according to the present embodiment. In the raster address conversion shown in the figure, the pixel width of the line buffer is 24. In this case,
It is sufficient for the line buffer to have a size of 8 × 24. Then, an 8 × 24 line buffer is replaced with a 1 × 8 pixel group (PIXELGROUP, hereinafter simply referred to as “P”).
It is divided into 24 groups by G '), and they are numbered from 0 to 23 as shown in FIG.

FIG. 7 is a diagram showing the sequence of PG generation during raster block conversion in this embodiment. In FIG. 7, the first column shows the sequence of PGs to be written, and the second column shows the sequence. The following shows the sequence of PGs in which reading and writing are performed with a delay of half a clock in one cycle. Therefore, a specific raster address conversion in this embodiment will be described.

In FIG. 5, if the pixel width of the line buffer is 24 and sub-sampling is not performed as described above, the number of blocks of one line (N
B) is assigned 3 and PG is set to 24. In the next step S2, since it is the first processing routine, increment (INCREMENT) = 1, predetermined value (VALUE) = 0, and number of groups (I) = 1. In step S3, since I is equal to or less than PG (= 24), the determination result is YES, so the process is performed in step S3.
Go to 4 and increment for VALUE. That is, PG address = 0 and VALUE = 1. Here, since PG address = 0, the first row and first column of the pixel groups for raster block conversion shown in FIG.

At step S5, VALUE (= 1)
Is not greater than or equal to PG (= 24), the determination result there is NO, and the process proceeds to step S6. Then, in step S6, it is determined whether or not I and NB are equal. Here, since I (= 1) is not equal to NB (= 3), the process proceeds to step S9. That is, I is incremented by 1 and I = 2.

Next, as a result of the above processing, the processing returns to step S3 again, and in this case as well, I (= 2) is PG (=
24) or less, the determination in step S3 is YES, and in the following step S4, PG address = 1, VALUE = 2. Here, since the PG address = 1, the first row, second column in FIG. 7 is 1.

In the following step S5, VALUE (=
Since 2) is not greater than or equal to PG (= 24), the determination is NO.
And the process proceeds to step 6. This step S6
Then, since I (= 2) is not equal to NB (= 3),
The determination is NO. Then, the process proceeds to step S9 to set I = 3. In the case of I (= 3), the determination in step 3 is YES because the value is PG (= 24) or less, and in the following step S4, PG address = 2 and VALUE = 3. Here, since the PG address = 2, the first row, third column becomes 2 in FIG. Further, in step S5, VALUE (= 3) is not greater than or equal to PG (= 24), so the determination is NO, and the process proceeds to step 6. In step S6, I (= 3) is NB
Since it is equal to (= 3), the determination is YES. As a result, in step S8, the next increment (NEXT_INCREMENT) =
3 Then, in the subsequent step S9, I = 4 as a result of the increment of I.

Thereafter, similar processing is executed by repeating the loop of steps S3 to S9, and as a result, the third, fourth, fifth, sixth, and subsequent rows in the first row of FIG.
7, 8, 9, 10, 11, ... 23. After the processing for the first line is completed, the value of I becomes 25.
In step S3 of, I (= 25) is equal to PG (= 2
4) Not less than Therefore, the determination in step S3 is NO, and the process is returned to step S2. As a result, increment = 3, VALUE = 0, and I = 1. In the determination at the subsequent step S3, I (= 1) is P
Since G (= 24) or less, YES is obtained, and in step S4, PG address = 0 and VALUE = 3. Here, since the PG address = 0, the second row and first column in FIG. 7 becomes zero.

At step S5, since VALUE (= 3) is not greater than or equal to PG (= 24), the determination is NO,
In the following step S6, I (= 1) is not equal to NB (= 3), and therefore the process proceeds to step S9. Therefore, I = 2
Becomes Next, when the process proceeds to step S3 again, I
Since (= 2) is less than or equal to PG (= 24), PG address = 3 and VALUE = 6 in step S4. Here, since the PG address = 3, the second of FIG.
The second column in the row is 3. Then, in step S5, VA
Since LUE (= 6) is not greater than or equal to PG (= 24), NO is determined. Then, in step S6,
Since I (= 2) is not equal to NB (= 3), I = 3 is set in step S9.

After setting I = 3, the process returns to step 3, so that the value of I is equal to or less than PG (= 24), and the determination there becomes YES. Then, in the subsequent step S4, PG address = 6 and VALUE = 9. The PG address = 6 here means that the second row, third column in FIG. 7 becomes 6. Further, in step 5, VALUE (= 9) is not greater than or equal to PG (= 24), so a NO determination is made, and in the next step S6,
I (= 3) is equal to NB (= 3), so the determination is YES.
become. As a result, in step 8, the next increment (NEXT_INCREMENT) =
9 Then, in step S9, I = 4.

Thereafter, by repeating the processing loop from step S3 to step S9 as described above, as shown in FIG. 7, after the fourth column of the second row, 9, 12, 15,
18 ... If I (= 8) is obtained in step S2 as a result of incrementing I, this I is PG
Since (= 24) or less, YES is determined in step S3. Then, in step 4, PG address = 21 and VALUE = 24. Here, since the PG address = 21, FIG.
The 2nd row and 8th column of becomes 21.

At step S5, VALUE (= 24)
Is greater than or equal to PG (= 24), the process proceeds to step S7, where VALUE = 1 and I (= 8) is not equal to NB (= 3) in the following step S6. Is NO. As a result, I = 9 in step S9. Next, in step S3,
Since I (= 9) is less than or equal to PG (= 24), it is determined to be YES, and in step S4, PG address = 1 and VALUE = 4. Here, since PG address = 1, 1 in the second row and ninth column in FIG.

At step S5, VALUE (= 4) is not greater than or equal to PG (= 24), so it is determined to be NO. At subsequent step S6, I (= 9) is not equal to NB (= 3). Then, in step S9, I = 10. Thereafter, the processing loop from step S3 to step S9 is repeated in the same manner, so that 4, 7, 10, ...

As described above, the raster block conversion is realized by repeating the above-described processing operation up to the final pixel of the image. However, at the time of the last 8-line raster block conversion, only reading is performed within one cycle and writing is not performed with respect to the line buffer. As described above, according to the present embodiment, a single line buffer is read with a certain clock at the same address, and a write operation is performed half a clock later to perform raster block conversion. By executing the address generation of, the number of line memories can be saved, and the image data input in raster sequence can be converted into block sequence in a pipeline manner.

Further, by changing the access sequence of the pixel data stored in the line buffer to halve the memory capacity for the image data of the same line, the cost of the apparatus can be reduced. In the above embodiment, the
As shown in FIG. 2, the line buffer unit 2 is arranged in the previous stage of the color conversion unit 3, but the present invention is not limited to this.
The line buffer unit 2 may be arranged between the color conversion unit 3 and the DCT unit 4. Further, in the above embodiment, the width of the input image is set to 24, but the width is not limited to this.

Further, in the above embodiment, the block size for raster block conversion is 8 × 8, but in step S1 of FIG. 5, PG = NB * 8 '8' is changed to an arbitrary value (n). By doing so, the height of the unit block can be changed, and regarding the pixel group,
By changing the value from 1 × 8 to 1 × m (m is an arbitrary number), it is possible to convert into an arbitrary block sequence of n × m blocks. [Second Embodiment] Next, the inverse raster block conversion performed when decompressing compressed data according to the second embodiment of the present invention will be described with reference to FIGS.

FIG. 8 is a flowchart showing this raster block inverse transformation, and FIG. 9 is a PI of the line buffer.
The number of XEL GROUP (PG) is shown. This line buffer is the same as that used for raster block conversion at the time of compression, and is composed of 24 groups of 1 × 8 pixel PGs, but the numbering of PGs is slightly different.

FIG. 10 shows the sequence of PGs for inverse raster block conversion. In FIG. 10, the first column shows the sequence of PGs written first, and the second and subsequent columns are shown. It shows a sequence of PGs in which reading and writing are performed with a half clock delay in one cycle. However, in the PG, the sequence is always performed from left to right.

In FIG. 8, in step S10,
Since the number of blocks (NB) on one line is 3, PG =
3 × 8 = 24, next increment = 1. In the next step S11, increment (INCREMEN
T) = 1, a predetermined value (VALUE) = 0, and the number of groups (I) = 1 are set. Then, in step S12, since I (= 1) is equal to or less than PG (= 24), the determination result is YES, and the process proceeds to step S13. Therefore, VALUE is incremented. That is, PG address = 0 and VALUE = 1. Here, since PG address = 0, FIG.
In the order of the pixel groups of the raster block inverse transformation shown in (1), the first row, first column becomes 0.

In step S14, VALUE (= 1)
Is not greater than or equal to PG (= 24), the determination result there is NO, and the process proceeds to step S16. Then, in step S16, since I (= 1) is not equal to 8, the determination is NO, and the process proceeds to step S18. That is, I is incremented by 1 and I = 2.

Next, as a result of the above processing, the processing returns to step S12, and in this case as well, I (= 2) is PG.
Since (= 24) or less, the determination in step S12 is YE.
Then, in step S13, PG address = 1 and VALUE = 2. Here, since PG address = 1, FIG.
The first row, second column of is 1.

In the following step S14, VALUE (=
Since 2) is not greater than or equal to PG (= 24), the determination is NO.
And the process proceeds to step 16. This step S
In 16, since I (= 2) is not equal to 8, the determination is NO. Then, the process proceeds to step S18,
Let I = 3. Hereinafter, similar processing is repeated by repeating the loop from step S12 to step S18, and if I = 8 in step S16, the determination is YES and the process proceeds to step S17 where the next increment = 8. Become. Therefore, from the third row onward in the first row of FIG. 10, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11,
…, 23.

By these processes, the data of the first 8 lines is written in the line buffer. Hereinafter, a method of generating an address of the line buffer in which reading and writing are performed in one cycle will be described. After PG address = 23, through steps S14, S16, S18, and in step S12, I (= 25) is changed to PG (=
24) Not less than. Therefore, the determination in step S12 is NO, and the process is returned to step S11. As a result, in step S11, increment = 8, VALUE = 0, and I = 1. The determination in the subsequent step S12 is YES because I (= 1) is less than or equal to PG (= 24), and the determination is step S1.
In 3, PG address = 0 and VALUE = 8. Here, since PG address = 0, FIG.
The second row, first column of is 0.

At step S14, VALUE (= 8)
Is not greater than or equal to PG (= 24), the determination is NO, and I (= 1) is not equal to 8 in the following step S16, so the process proceeds to step S18, resulting in I = 2. Next, when the process proceeds to step S12 again, I
Since (= 2) is less than or equal to PG (= 24), the following step S
At 13, PG address = 8 and VALUE = 16. Here, since PG address = 8, FIG.
The second row, second column of is 8.

At the next step S14, VALUE (=
16) is not more than PG (= 24), so N
O is judged. Then, in step S16, I (=
2) is not equal to 8, so in step S18, I =
Set to 3. After that, the process returns to step 12, and the value of I (= 3) is equal to or less than PG (= 24).
The determination there becomes YES. And the following step S
In 13, PG address = 16 and VALUE = 24. The PG address = 16 here is the second one in FIG.
This means that the third column in the row becomes 16. In step 14, VALUE (= 24) is equal to or larger than PG (= 24), so a YES determination is made, and in step S15, VALUE = 24-24 + 1 = 1.

Next, in step S16, since I (= 3) is not equal to 8, the determination is NO. As a result, in step 18, I = 4. The process returns to step S12 again, and I (= 4) is equal to or less than PG (= 24), so the determination is Y.
The process becomes ES, and the process proceeds to step S13. And
PG address = 1 and VALUE = 9. here,
Since PG address = 1, 1 in the second row and fourth column in FIG.
Becomes

Next, in step S14, VALUE
Since (= 9) is not greater than or equal to PG (= 24), NO is determined and the process proceeds to step S16, where I (= 4) is not equal to 8, so the determination is NO. Then, I = 5 in step S18, and the process returns to step S12. Thereafter, by repeating the processing loop from step S12 to step S18 in the same manner, the second line, fifth line in FIG.
From the column onwards, 9, 17, 2, 10, 18, 3, ..., 23
Becomes

In this way, the raster block inverse transform is realized by repeating the above-described processing operation up to the final pixel of the image. However, at the time of the last raster block inverse conversion of 8 lines, only reading is performed within one cycle and writing is not performed with respect to the line buffer. The present invention may be applied to a system including a plurality of devices or an apparatus including a single device. Further, it goes without saying that the present invention can be applied to the case where it is achieved by supplying a program to a system or an apparatus.

As described above, according to the present embodiment, by writing and reading the image data according to the addresses indicating the writing order and the reading order of the image data determined for the n × k pixel groups, A single memory can be used to convert image data input in raster order into image data in block order, or to convert image data in block order into image data in raster order. The cost of the device can be reduced by using the memory. Furthermore, by making the integers n, m, and k indicating pixel groups variable, it is possible to perform raster block conversion into blocks of arbitrary size.

[0043]

As described above, according to the present invention,
A single line buffer can be used to perform raster block conversion of image data, and the cost of the device can be reduced.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a raster block conversion unit in an image processing apparatus according to a first embodiment of the present invention.

FIG. 2 is a diagram showing raster-sequential image data.

FIG. 3 is a diagram showing block-sequential image data.

FIG. 4 is a diagram showing a configuration of a conversion unit related to data decompression in the embodiment.

FIG. 5 is a flowchart showing a raster block conversion procedure according to the embodiment.

FIG. 6 is a diagram showing pixel group numbers of a raster block conversion line buffer according to an embodiment.

FIG. 7 is a diagram showing a sequence of pixel groups for raster block conversion according to an embodiment.

FIG. 8 is a flowchart showing a raster block inverse conversion procedure according to the second embodiment.

FIG. 9 is a diagram showing PG numbers of a line buffer according to the second embodiment.

FIG. 10 is a diagram showing a sequence of pixel groups for raster block inverse transformation according to the second embodiment.

[Explanation of symbols]

1 Address generator 2,15 Line buffer section 3 color converter 4 DCT section 5 Quantizer 6 Encoding section 11 Decoding unit 12 Dequantizer 13 Inverse DCT section 14 Reverse color converter

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI // H03M 7/30 H03M 7/40 7/40 H04N 7/133 Z (72) Inventor David Reid New South Wales, Australia 2074, Tsuramura, Denman Street 11 (56) References JP-A-64-20786 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) H04N 1/41 H04N 1/21 H04N 7 / 30

Claims (4)

(57) [Claims] 1. Reading and writing image data of a single line buffer having a storage size for a predetermined number of lines (n lines) is performed with a predetermined size (m pixels) smaller than the storage size for one line of the line buffer. An image processing method for performing raster block conversion on image data by controlling in units of data groups, wherein image data input in raster order or in a block order of m × n pixels is stored by the storage size of the line buffer, In the data group unit, write to the line buffer, the image data held in the line buffer is read in the data group unit in a block order or raster order of m × n pixels, immediately after the reading, Input / output to / from the read address of the line buffer for writing the image data to be input next in units of the data group.
Repeated until there is no image data to be stored, the line buffer held in the line buffer
Image data for the memory size of
Or, read in the data group unit so that the data will be in raster order.
Image processing method, characterized in that to. 2. The image processing method according to claim 1, wherein the writing is performed with a half clock delay from the reading timing. 3. A single line buffer having a storage size for a predetermined number of lines (n lines), and a predetermined size for reading and writing image data of the line buffer, which is smaller than the storage size for one line of the line buffer. (M pixels) has a conversion unit for performing raster block conversion on the image data by controlling the data group as a unit, and the conversion unit is image data input in a raster order or a block order of m × n pixels. Is stored in the line buffer
For each size, the data group is written in the line buffer, and the image data held in the line buffer is read in the data group unit in a block order of m × n pixels or a raster order, and the reading is performed. Shortly after,
Reading / writing of the line buffer for writing the image data input next to the read address in the unit of the data group
And repeated until the image data to be input is not, before
Of the line buffer held in the line buffer
Image data for the memory size is sorted in m × n pixel block order.
Alternatively, the image processing apparatus is characterized in that the data is read in the data group unit in raster order . 4. The image processing apparatus according to claim 3, wherein the control unit performs the writing with a half clock delay from the reading timing.