JP2001157039A - Resolution converting method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a resolution converting method for reducing loads on hardware and for speedily reducing the resolution of inputted image data. SOLUTION: This resolution converting method is constituted of plural steps comprising a resolution conversion step 8a for reducing the resolution of inputted image data [RGB data (Ri, Gi and Bi)]. The reduction in the resolution for the resolution conversion step 8a is attained through addition and an approximate operation using bit shift and the correction of the approximate operation is performed concurrently in a step subsequent to the resolution conversion step 8a together with the other operation.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a resolution conversion method capable of rapidly reducing the resolution of image data while reducing the burden on hardware.

Conventionally, image data (RGB data) input from a scanner is converted into a resolution conversion step (resolution reduction step), a γ conversion step, an RGB conversion step,
There is known a resolution conversion method for performing resolution conversion through each step called an RG conversion step (see FIG. 5).
For example, RGB data (Ri, Gi, Bi) input at a resolution of 300 dpi × 300 dpi is converted to 50 dpi ×
50 dpi IRG data (I, Ro, Go)
First, as shown in FIG. 4, in the X direction (main scanning line direction) and the Y direction (sub scanning line direction), as shown in FIG. Into one dot. Specifically, the RGB data of the odd-numbered line (Y direction) is discarded, the RGB data of the even-numbered line (18 data for three lines) is added, and the added value is divided by 18 to obtain an average. A value is obtained (see the formula in FIG. 5) and the average value (_R, _
G, _B) are the values of the RGB data of one dot after the resolution reduction. Hereinafter, data is processed for each dot after the resolution reduction. That is, the RGB data after the resolution reduction is
The color distortion is removed by the γ conversion step, the scanner characteristics are corrected by the RGB conversion step, and the IRG
The RGB data is converted to the IRG data (I,
Ro, Go). Here, "I" is the luminance (Inte
nsity).

[0003] The RGB conversion and the IRG conversion are each performed by matrix calculation (see the mathematical formula in Fig. 5).

[0004] The data (I,
Ro, Bo) are used for a bill recognition system (a system for preventing copying of bills) of a color copying machine.

[0005]

In the above-described resolution conversion method, it is necessary to obtain an average value by dividing the added value by 18 in order to reduce the resolution of the input RGB data. However, performing such a division imposes a heavy burden on hardware and requires a long time for the operation. In particular, when the number of operations is enormous, the load on the hardware becomes extremely large, and the processing time becomes long. In order to shorten the processing time, it is conceivable to use a high-performance CPU, but this is not preferable because it leads to an increase in cost.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a resolution conversion method capable of rapidly reducing the resolution of input image data while reducing the burden on hardware.

[0007]

Means for Solving the Problems In order to solve the above problems, the present inventors have diligently made efforts to complete the present invention. That is, the resolution conversion method of the present invention includes a plurality of steps including a resolution conversion step of lowering the resolution of input image data. Then, the resolution reduction in the resolution conversion step is performed by an approximate operation using addition and bit shift, and the correction of the approximate operation is performed together with other operations in a step subsequent to the resolution conversion step. Features.

According to this configuration, the average value is obtained by an approximation operation using addition and bit shift, so that the burden on hardware is small. In addition, since the correction of the approximation calculation is performed in a subsequent step together with the calculation processing performed in the step, the load on the hardware does not increase.

Further, in the present invention, the plurality of steps including the resolution conversion step may include a resolution conversion step of reducing the resolution of the RGB data input as the image data, and a gamma conversion of the reduced resolution RGB data. Converting the γ-converted RGB data to R
An RGB conversion step of performing an RGB conversion, and an IRG conversion step of performing an IRG conversion of the RGB converted RGB data, and the resolution conversion step integrates 6 dots × 6 dots of the input RGB data into one dot. 300dpi x 300d
This is a step of lowering the resolution from 50 dpi to 50 dpi × 50 dpi, and this resolution conversion step discards 18-dot RGB data in the 6-dot × 6-dot matrix of the input RGB data, and In the case where the resolution is to be reduced by adding RGB data for 18 dots and dividing the added value by 18 to obtain an average value, the following operation is performed.

In the resolution conversion step, the approximate operation is an approximate value of 1/6 (1 + 1 /
4) Multiplying by / 8 and further multiplying by ３ in a step after this resolution conversion step. Then, the approximation calculation is corrected together with the multiplication by 1/3.

According to this configuration, dividing the added value by 18, that is, multiplying the added value by 1/18 is performed by multiplying the added value by (1/6) × (1/3). The calculation of multiplying by 1/6 of the former is performed in the resolution conversion step (the "resolution reduction step" in the embodiment of the present invention). The latter operation of multiplying by 1/3 is performed in a step subsequent to the resolution conversion step, together with the operation performed in this step.

However, 1/6 is an approximate value thereof (1+
1/4) / 8 is used. This (1 + 1/4) / 8 is 1
This is an approximate value of 1/6 smaller than / 6. According to this operation, the operation can be performed by bit shift and addition. Incidentally, 1/4 can be operated by shifting two bits to the right, and 1/8 can be operated by shifting three bits to the right. In addition, when multiplying by 1/3, the approximation calculation is also corrected to reduce the number of calculations. This reduces the load on the hardware and shortens the calculation time.

In the present invention, it is preferable to perform the RGB conversion step and the IRG conversion step together. Further, at this time, it is preferable that the calculation of multiplying by 1/3 and the correction of the approximate calculation are also performed together. That is, in order to reduce the number of times of calculation, calculation of multiplying by 1/3, correction of approximation calculation (correction required by multiplying by an approximate value of 1/6 instead of multiplying by 1/6), RGB conversion, IRG
It is preferable that the four operations of conversion be performed by one linear conversion ("Afin conversion" in the embodiment of the invention).

By doing so, the operation can be performed in a short time without putting a burden on the hardware.

Further, the present invention relates to the input RGB
When the resolution of the data is 600 dpi × 600 dpi, every other data in the X and Y directions of the RGB data is discarded, and 300 dpi ×
A resolution conversion method characterized in that the resolution conversion step is performed after conversion into RGB data having a resolution of 300 dpi.

According to this, the image is reliably converted to 300 dpi.
The resolution can be reduced to × 300 dpi. After the resolution is reduced, the image is handled as a normal 300 dpi × 300 dpi image.

[0017]

Embodiments of the present invention will be described below in detail with reference to the drawings. In the present embodiment, a case where the resolution conversion method of the present invention is applied to a bill recognition system of a color copying machine will be described.

{Configuration of Banknote Recognition System} First, the configuration of a banknote recognition system to which the resolution conversion method according to the embodiment of the present invention is applied will be described (see FIG. 2).

FIG. 2 is a block diagram showing a configuration of a bill recognition system and a schematic configuration of a color copier. The bill recognition system 1 is an ASIC (Application Specific I / O).
integrated circuit) 2, SRAM (Static Random Acc)
ess Memory) 3, MC (Micro Computer) 4 and DRA
An M (Dynamic Random Access Memory) 5 is included. The color copier further includes a scanner SC, a host computer HC, and a printer PR.

An image read from the scanner SC of the color copying machine is input to the ASIC 2 as RGB data. Then, the ASIC 2 determines whether or not the image read by the color scanner SC is a banknote (whether or not a banknote is included in the read image) and transmits the information to the host computer HC of the color copying machine. The host computer HC prohibits copying when the read image is determined to be a banknote, and when the read image is determined to be not a banknote, the host computer HC uses the color printer PR to execute the image read by the scanner SC. Is output and color copying is performed.
Note that the RGB data includes red (Red) and green (Green) of each pixel.
n), blue data and its resolution is 300
The resolution is 300 dpi, and each data of R, G, and B is 8-bit data. Note that the resolution and RG
The B data is not limited to the numerical values described above.

In this embodiment, the X direction (main scanning line direction) is the horizontal direction of an image read by the scanner SC, and the Y direction (sub scanning line direction) is the same vertical direction.

ASIC2 is ARBITOR6, PIX
_FIFO (First In First Out) 7, resolution conversion unit (hereinafter referred to as “RSU”) 8, RSFE_FIFO
9, a feature extraction unit (hereinafter referred to as "FEU") 10,
FEPM_FIFO 11, pattern matching unit (hereinafter referred to as “PMU”) 12, PMMC_FIFO 1
3, an MC interface (hereinafter, referred to as “MC_IF”) 14, a control unit (hereinafter, referred to as “CU”) 15, a ROM (Read Only Memory) 16, and the like.

ARBITOR 6 is composed of SRAM 3 and RSU
8, arbitrate access between the FEU 10, the PMU 12, and the MC_IF 14. When the SRAM 3 is simultaneously accessed from the RSU 8, FEU 10, PMU 12, or MC_IF 14, the access is made in accordance with a preset priority.

The PIX_FIFO 7 is a FIFO that transfers RGB data from the scanner SC to the RSU 8.

The RSU 8 converts the RGB data from the scanner SC into resolution conversion, γ conversion and Afin conversion, and writes the converted IRG data into the SRAM 3. In the resolution conversion, the resolution of the RGB data input from the scanner SC is changed from 300 dpi × 300 dpi to 50 dpi × 50 d.
pi to reduce the load on each subsequent process.
After the resolution conversion, each of the R, G, and B data becomes 10-bit data. In the Afin transformation, the RGB transformation, the IRG transformation, and the correction of the approximation operation are performed by one linear transformation. Incidentally, the IRG data after Afin conversion is luminance (Intensity), red, and green data of each pixel. Each of I, R, and G data is 8-bit data. The Afin transform will be described later in detail.

The RSFE_FIFO 9 is a FIFO that transfers the IRG data converted by the RSU 8 to the FEU 10 as a command. The IRG data is stored in the SRAM
3 and the IRG data of the SRAM 3 is delivered to the FEU 10 by the address.

The FEU 10 converts the IR converted by the RSU 8
A feature amount for performing pattern matching is extracted from the G data, and the extracted data (hereinafter referred to as “FE data”)
Is written to the SRAM3. Feature extraction is 40 dots x 4
The process is performed in units of one block, with 0 dot as one block. Further, one row in the X direction is defined as one unit, and feature extraction is performed for each block while shifting each dot in the X direction by 5 dots (each block adjacent in the left and right direction has a portion overlapping by 35 dots). When one unit is completed, the feature of the next unit is extracted by shifting it by 5 dots in the Y direction (each vertically adjacent block has an overlapped portion of 35 dots). FIG. 3 is a diagram for explaining the concept of resolution conversion and the concept of a block.
Input image data (RGB data) having a resolution of i.times.300 dpi (upper diagram in FIG. 3) is 50 dpi in RSU8.
The image data is converted into low-resolution image data of 50 dpi and also converted to IRG data. And 40 dots x
The feature amount is extracted in the FEU 12 with 40 dots as one block (lower diagram in FIG. 3). The resolution conversion will be described later.

It should be noted that the FE data is composed of one block 1600.
800th level value (1
, “Median value”), I (luminance) data, representative values (20 × 2), I (luminance), (R−I) which are addition data for every two dots in the X and Y directions. )Color difference,
(GI) The variance (3 × 2) of the addition data for each two dots in the X and Y directions of the color difference.

The FEPM_FIFO 11 is a FIFO that transfers FE data extracted for each block in the FEU 10 to the PMU 12 that performs pattern matching as a command. Note that the FE data is written in the SRAM 3 and the PMU 12 stores the FE data by the address.
The FE data in the RAM 3 is delivered. The unit of data to be transferred is one bucket unit composed of a plurality of blocks.

The PMU 12 performs pattern matching between the FE data for each block extracted by the FEU 10 and each banknote data in the banknote database stored in the SRAM 3. SRAM)
Write to 3. In the pattern matching, the median value and the variance value in the FE data are screened by banknote data, and the FE data that has passed the screening is calculated and compared with the representative value of the FE data (the similarity is calculated in advance). When it exceeds a predetermined threshold value, it is determined that there is matched data, and MC4 is subjected to structure matching). Pattern matching is
This is performed for each block in which the feature extraction has been performed, and a pattern match between the input image and the banknote is determined for each block.

The PMMC_FIFO 13 is a FIFO for transferring PM data matched by the PMU 12 to the MC 4 as a command. The PM data is written in the SRAM3, and the MC4 is stored in the SRAM by an address.
3 PM data is delivered.

The MC_IF 14 is an interface for accessing the SRAM 3 by the MC 4.

The CU 15 is an interface and a ROM for the MC 4 to read the setting and status of each unit.
16 and performs parallel communication with the host computer HC. The CU 15 receives a command from the host computer HC through lbus. The host computer HC obtains data and status by reading a read register (not shown).

The ROM 16 stores an RO for the monitor program.
M. The monitor program is executed by the host computer H
Communication with C and management of minimum hardware. The program of the bill recognition system 1 is downloaded from the host computer HC by the function of the monitor program.

The SRAM 3 stores 50 dpi from the RSU 8.
The FE data from the FEU 10 is stored in the FEBuff in the ImageBuffer area (not shown).
In the er area (not shown), the PM data from the PMU 12 is stored in the PMBuffer area (not shown), and the banknote database required for pattern matching of the PMU 12 is Data.
It is stored in the Base area.

The MC 4 performs overall control of the banknote recognition system 1, communication with the host computer HC, and structure matching on the pattern matching result of the PMU 12. In the structure matching, three pieces of PM data satisfying a predetermined condition are selected from the PM data, and a distance between the other two blocks 3 and an angle between the two blocks 3 are determined based on one of the three blocks. Is calculated and compared with each data of the stored structure matching database. If the distance and the angle match, it is determined that the image read from the scanner SC corresponds to a bill (it is determined that a bill is included).

The DRAM 5 stores a program downloaded from the host computer HC and a database for structure matching. Also, it is a work area of MC4. This is also an area where all the lists of the results of the pattern matching of the PMU 12 are created.

{Resolution Conversion Method} The above is an outline of the bill recognition system to which the resolution conversion method according to the embodiment of the present invention is applied. Hereinafter, the resolution conversion method according to the embodiment of the present invention will be described in more detail. In addition,
The resolution conversion method according to the embodiment of the present invention is performed in the RSU 8 of the bill recognition system 1 shown in FIG.

The RSU 8 converts the resolution of the input RGB data, and converts the RGB data having a resolution of 300 dpi × 300 dpi into IRG data having a resolution of 50 dpi × 50 dpi. Figure 1 shows the resolution conversion.
As shown in the block diagram of FIG.
This is performed through the γ conversion step 8b and the Afin conversion step 8c. Note that the resolution conversion here includes not only lowering the resolution but also the original RGB data (Ri,
Gi, Bi) to IRG data (I, Ro, Go).

[Low-resolution] Low-resolution step 8a
Converts the input RGB data (Ri, Gi, Bi) having a resolution of 300 dpi × 300 dpi to 50 dpi × 50.
The resolution is reduced to dpi RGB data (_R, _G, _B).

In this embodiment, the resolution is reduced (see FIG. 4).
In the X direction (main scanning line direction) and the Y direction (sub scanning line direction), a total of 36 dots of 6 dots × 6 dots are reduced to one dot in the resolution reduction step 8a. Specifically, the RGB data of the odd-numbered line (Y direction) is discarded, and the RGB data of the even-numbered line (18 data for three lines) is added. Then, the respective average values (_R, _G, _B) are obtained. However, unlike the conventional example, when calculating the average value, the added value is not divided by 18, but multiplied by an approximate value of 1/6 (note that the correction of this approximate calculation and the calculation of multiplying by 1/3 will be described later). Afin conversion step 8c).

The operation in the resolution lowering step 8a is as follows.

[0043]

(Equation 1)

In the equations (1) to (3), (1 + ／) /
8 is an approximate value of 1/6. This calculation can be performed by a bit shift and one addition. (1 + 1/4) / 8 is 1
Although the value is smaller than / 6, by using such a small value, the operation can be performed without overflow. Incidentally, the operations of the expressions (1) to (3) are performed as follows (see FIG. 4).

An added value for 6 dots of the first line (line # 0) is obtained, multiplied by (1 + ／), and further multiplied by ８.

An additional value for 6 dots of the second line (line # 2) is obtained, multiplied by (1 + ／), and further multiplied by ８. The calculated value is added to this calculated value.

The sum of 6 dots of the third line (line # 4) is obtained, multiplied by (1 + 1/4), and further multiplied by 1/8. The calculated value is added to this calculated value.

When the resolution of the input RGB data is 600 dpi × 600 dpi, the data is discarded every other data in the X direction and Y direction, and 300 dpi × 300 dpi.
The processing in the subsequent steps is performed as RGB data having a resolution of 300 dpi.

The distinction between 600 dpi and 300 dpi is determined by RX and RY of the command register (ResCmd0). The X direction is designated by RX and the Y direction is designated by RY. When the value is 1, the resolution is 600 dpi, and when the value is 0, the resolution is 300 dpi. That is, 600 dpi × 300 dpi or 30 dpi
It is also possible to convert RGB data input at a resolution of 0 dpi × 600 dpi to a resolution of 300 dpi × 300 dpi. By writing this command register, initialization for inputting one screen is performed, and FE
An initialization command (INIT) is sent to U10, and the subsequent RGB input is processed. After the input of the RGB data of one screen is completed, the command register is written and initialized to input the RGB data of the next screen.

The intermediate result during the operation from line # 0 to line # 5 (see FIG. 4)
Uses 1 Line in the image buffer area (not shown). The wait time required for reading / writing from / to the ImageBuffer area is set to allow a maximum of three clocks after issuing a request signal to ARBITOR6. The bit length of the intermediate result RGB data is 10
Is a bit.

One line of resolution conversion (line # 0 to l
When the operation up to ine # 5 is completed, resolution conversion for the next one line is performed. In the Line where the intermediate result in the ImageBuffer area is stored, the final result (I, Ro, Go) obtained by the Afin conversion is stored.

The number of pixels in the X direction in the input RGB data is represented by PixPerLine and is set in a register. PixPerLine has a resolution of 30
Even in the case of 0 dpi × 300 dpi, it may not always be a multiple of 6. Similarly, the resolution is 600 dpi x 60
Even at 0 dpi, it may not always be a multiple of 12. If there is a fraction, the fraction is discarded without conversion (see FIG. 4). On the other hand, 50dp with reduced resolution
The number of dots in the main scanning line direction of i × 50 dpi data (_R, _G, _G) is represented by DotPerLine and set in a register. Incidentally, one of the following is set in the register in accordance with the resolution of the input RGB data (Int is a function for converting to an integer). DotPerLine = Int (PixPerLine / 6) ← 300 DotPerLine = Int (PixPerLine / 12) ← 600 [γ-conversion] In the γ-conversion step 8b, the R after the resolution reduction is set.
Gamma conversion is performed on the GB data (_R, _G, _B) to remove color distortion. The calculation in the γ conversion step 8b is shown below. Note that the powers in the expressions (4) to (6) are (1/1.
8).

[0053]

(Equation 2)

[Afin Conversion] Afin Conversion Step 8
In c, the RGB conversion and the IRG conversion are performed together, but further, the correction of the approximation calculation in the resolution lowering step 8a and the calculation by multiplying by 1/3 which is not performed in the resolution lowering step 8a are also performed. In the present embodiment, the calculation in the Afin conversion step 8c is performed by one linear conversion. As a result, the number of operations can be reduced. The input data (Rt, Gt, Bt) has 10 bits each, and the output data (I, Ro, Go) has 8 bits each.

The operation in the Afin conversion step 8c is as follows.

(Equation 3)

The fractional part on the right side of the equation (7) is used for correcting the calculation of multiplying by 1/3 which was not performed in the resolution lowering step 8a and the approximation calculation in the resolution lowering step 8a (FIG. 1). As a result, the sum of the RGB data in the resolution reduction step 8a (18 data for 3 lines) is multiplied by 1/18 as a result. In other words, (1/18) becomes (1 + ／) / 8
When multiplied by / [(1 + 1/4) / 8], it becomes 1/18.

The 3 × 3 matrix portion on the right side of the right side of equation (7) is for performing RGB conversion (FIG. 1).
reference). The characteristics of the scanner SC are corrected by the RGB conversion.

Similarly, the 3 × 3 matrix portion on the left side of the right side of equation (7) is for performing IRG conversion (see FIG. 1).

Note that the actual calculation is performed not by the expression (7) but by the expression (8) in order to further reduce the load on the hardware. Cij 'in equation (8) is a setting register, and the calculated value is set in the Afin transform coefficient register in advance (Cij' is a 10-bit two's complement number). Also, (8)
1/1024 in the equation is a coefficient for converting Cij 'to an integer, and corresponds to extracting the operation result by shifting it by 10 bits.

According to the resolution conversion method configured as described above, the input RGB data can be processed quickly and converted to IRG data of a predetermined resolution without imposing a burden on hardware.

The present invention described above is not limited to the above-described embodiment of the invention, but can be appropriately modified and implemented within the scope of achieving the objects and effects of the present invention.

For example, the resolution of the input RGB data is 300 dpi × 300 dpi or 600 dpi × 600.
It is not limited to dpi. Further, the resolution after the reduction is not limited to 50 dpi × 50 dpi. Furthermore, as a means for lowering the resolution, 6 dots × 6
There is no limitation to combining dot data into one dot. Further, the resolution conversion method of the present invention can also be applied to audio data.

[0063]

According to the present invention, since the operation is performed by using the bit shift and the addition, the load on the hardware is reduced. Further, the correction of the approximation calculation is performed in a step subsequent to the step of performing the approximation calculation together with the calculation performed in the step, so that the number of calculations can be reduced. Therefore, the burden on hardware can be further reduced. At the same time, the calculation time is shortened.

Even if the resolution of the input RGB data is 300 dpi × 300 dpi, 600 dpi × 60
Even at 0 dpi, resolution can be reliably converted to 50 dpi × 50 dpi IRG data by using bit shift and addition without burdening hardware and reducing the number of operations.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a procedure of a resolution conversion method according to an embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration of a bill recognition system to which a resolution conversion method according to an embodiment of the present invention is applied and a schematic configuration of a color copier;

FIG. 3 is a diagram illustrating the concept of resolution conversion and blocks in the bill recognition system of FIG. 2;

FIG. 4 is a diagram for explaining the resolution reduction in the resolution conversion method according to the embodiment of the present invention (also a diagram for explaining the resolution reduction in the conventional example).

FIG. 5 is a block diagram showing a procedure of a conventional resolution conversion method.

[Explanation of symbols]

 1 Bill Recognition System 8 RSU (Resolution Conversion Unit) 8a Low Resolution Step 8b γ Conversion Step 8c Afin Conversion Step

Continued on the front page F term (reference) 5B057 AA11 AA20 BA02 CA01 CA08 CA12 CA16 CB01 CB08 CB12 CB16 CC02 CD01 CD05 CE16 CH08 DA17 DC32 5C076 AA22 AA40 BB07 BB14

Claims (3)

[Claims] 1. A resolution conversion method comprising a plurality of steps including a resolution conversion step for lowering the resolution of input image data, wherein the resolution conversion step uses addition and bit shift. A resolution conversion method, wherein the correction of the approximate calculation is performed in a step subsequent to the resolution conversion step together with another calculation. 2. A plurality of steps including the resolution conversion step include a resolution conversion step of reducing the resolution of the RGB data input as the image data, and a gamma conversion step of gamma converting the reduced resolution RGB data. To convert the γ-converted RGB data to RGB
A GB conversion step, and an IRG conversion step of performing an IRG conversion of the RGB converted RGB data, and the resolution conversion step includes a step of combining the input RGB data of 6 dots × 6 dots into one dot to thereby achieve a resolution. 300 dpi x 300 dpi to 50 dpi
× 50 dpi, and the resolution conversion step is performed by the input RG
In the matrix of 6 dots × 6 dots of B data, 18
If the RGB data for the dots is discarded, the RGB data for the remaining 18 dots is added, and the added value is divided by 18 to obtain an average value, the resolution is reduced. In the resolution conversion step, by multiplying the sum by (1 + 1/4) / 8 which is an approximate value of 1/6 in the resolution conversion step, and further multiplying by 1/3 in a step subsequent to the resolution conversion step. 2. The resolution conversion method according to claim 1, wherein the correction of the approximation operation is performed together with the multiplication by 1/3. 3. If the resolution of the inputted RGB data is 600 dpi × 600 dpi, the RGB data
3. The resolution conversion step according to claim 2, wherein in the X direction and the Y direction of the B data, the data is discarded every other data, and the data is converted into RGB data having a resolution of 300 dpi × 300 dpi, and then the resolution conversion step is performed. 4. Resolution conversion method.