JPS62221076A - Picture processor - Google Patents

Abstract

PURPOSE:To shorten the processing time with a space filter by providing a conversion table where the output density value is decided to the density value of each picture element of an input picture based on the coefficient decided by the space filter parameter. CONSTITUTION:The picture data (input density data) equivalent to 9 picture element in all are simultaneously inputted to a coefficient table (RAM) 15 from a shift register 10 in the form of address signals. The tale 15 gives an access to the data in the address shown by the input density data and outputs this access data as the output density data. Then 9 pieces of output density data set by each coefficient parameter are inputted to an adder 13. The adder 13 obtains the sum total S and the bias value BIAS is added to the total S. As a result, a sum of products calculation is given to higher 8 bits out of the addition data and the normalized picture data is delivered. This picture data are written to a prescribed address of a picture memory.

Description

[Detailed Description of the Invention] [Configuration of the Invention] (Industrial Application Field) The present invention is applicable to, for example, ITM/(industrial use television)
The present invention relates to an image processing apparatus that uses a camera to inspect, for example, the external shape of an industrial product, and particularly relates to an image processing apparatus that has improved processing means for spatial filter processing that mainly uses product-sum calculations.

(Prior art) An image processing device used as a visual inspection device in a factory inspection line, etc. converts an image (analog image) from an ITV camera into a digital image, and performs spatial filtering processing, etc. on this digital image. to obtain a desired image pattern of the object to be inspected.

Here, what is required of this type of visual inspection apparatus is real-time performance, that is, the ability to process images input by a camera at a speed equal to or higher than that of human image processing.

Here, consider spatial filtering processing in an image processing device. That is, the spatial filtering process performs the calculation given by the following equation (1).

Here, Xa+, n, Ym, and n respectively indicate the values before and after spatial filtering in the density value of the pixel in m rows and n columns on the screen, as shown in FIG. Also, f 1
．． j is a loading coefficient and H is a normalization coefficient.

This spatial filter converts the value of the output image Ym,n into the weight (f i,j
) is determined by calculation, and this fI.

By selecting the j value, the characteristics of the spatial filter can be changed in various ways.

Conventionally, an image processing apparatus that uses a spatial filter as preprocessing of an image has a configuration shown in FIG. 4. That is, in FIG. 4, reference numeral 1 inputs a video signal from an ITV camera (not shown), converts it A/D, and converts it into an 8
Image human resources department that generates bit digital image data, 2
3 is an image memory that stores the image data in frame units; 3 is a spatial filter unit that removes noise from the image data from image memory 2; and 4 is a unit that stores the processing results of spatial filter processing unit 3 in frame units. It is an image memory. The image data stored in the image memory 4 is then subjected to various image processing to generate a desired image pattern, which can be used, for example, as inspection data to determine whether the object is a good product or a defective product. It has become.

Here, the spatial filter section 3 is constituted by the hardware shown in FIG. In Figure 6, 10 is, for example, 8
3 shift registers for inputting bit image data to configure a 3×3 matrix; 11 is 2 delay circuits for configuring a 3×3 matrix; 12 is a weight (f 1.j ) and each 9 multipliers that perform multiplication with pixels; 13 is an adder that calculates the sum of the 9 multiplication results;
14 is a divider for normalizing the addition result from the adder 13.

Next, the operation of the conventional spatial filter processing section configured as described above will be explained. That is, the image data read from the image memory 2 is transferred pixel by pixel to the shift register 10.
is input. Further, the image data that has passed through the delay circuit 1 is also input pixel by pixel to the shift register 10 at the next stage.

Here, the delay time of the delay circuit 11 is one horizontal scanning time in the ITV camera. Image data corresponding to a total of nine pixels is simultaneously input from the shift register 10 to the corresponding multiplier 12. In the multiplier 12, the weight (f i
, j), and the result is input to the adder 13.

The adder 13 takes the sum of the nine products (product-sum calculation result) S, and further adds a bias value BIA to the sum (product-sum calculation result) S.
S (for example, 1020) is added and the result is input to the divider 14.

This divider 14 converts this S+BIAS into a normalization coefficient H
It is normalized (result becomes 8 bits) by dividing by (for example, 8). The normalized image data is written to the image memory 4.

For example, input image data shown in FIG.
When processed using the spatial filter parameters shown in the figure, output image data (output density) shown in FIG. 7(b) is obtained.

For example, Y2,2 in FIG. 7(b) can be found using the following calculation for X2,2 in FIG. 7(a). That is, f(IOXI)+ (20X2)+ (IOXI)+
(20X(1)+ (10XO)+ (20XO)+
(100x (-1) ) + (100x (-2)
)+ (100X (-1))+10201 /8-8
Also, numbers below the decimal point are rounded down.

(Problems to be Solved by the Invention) The spatial filter section in the conventional image processing apparatus configured as described above has the following problems. That is,
Since the multiplier 12 and the adder 13 are used to perform the product-sum operation for each pixel, and the divider 14 is used to perform the normalization operation, the calculation takes a long time. For this reason, for example, spatial filtering cannot be performed at a speed higher than the same speed as the input of a video signal from an ITV camera. This is contrary to the real-time performance required when the image processing device is applied to a visual inspection device.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide an image processing device that can shorten the processing time of a spatial filter.

[Structure of the Invention] (Means for Solving the Problems) The present invention is characterized by taking the following measures in order to solve the above problems and achieve the object. That is, in an image processing apparatus having a spatial filter processing unit that performs a product-sum operation and normalizes each pixel of a target image to be image-processed using a coefficient based on a spatial filter parameter, the spatial filter processing unit A conversion table is provided that determines the output density value for the density value of each pixel of the human image based on coefficients based on parameters, the output pixel is determined for each pixel of the input image, and each of the output pixels is added. It is characterized by having a structure that allows

(Operation) By taking such measures, the product-sum operation and the normalization operation are realized using a conversion table without performing multiplication processing and division processing, resulting in faster processing. I can do it.

(Example) An example of the present invention will be described below with reference to FIGS. 1 to 3. In FIGS. 1 to 3, the same parts as in FIGS. 4 to 8 are given the same reference numerals, and their explanations are omitted. In FIG. 1, the multiplier 12 and divider 14 in FIG. A RAM (Random Access Memory) 15 is used to replace the function of .

Each address in the RAM 15 stores 16-bit output density data, and the manual density data output from each shift register 10 acts on the RAM 15 as an address signal.

Here, if the spatial filter parameters are of a 3×3 matrix as shown in FIG.
AM15 is the table shown in FIG. 2(a), with a coefficient of 1
The RAM 15 as a coefficient is the table shown in FIG. 2(b), the RAM 15 as a coefficient 0 is the table shown in FIG. 2(C), and the RAM 15 as a coefficient -1 is the table shown in FIG.
d), with RAM1 as a factor of -2.
5 is the table shown in FIG. 2(e), and the human power concentration is 8.
This is a bit address signal, and the output density data of each table, that is, the data stored in each address of the RAM 15, is set corresponding to the coefficient of the parameter, for example, as shown in FIGS. 2(a) to (e). This data indicates the 16-bit output density.

Next, the operation of this embodiment configured as described above will be explained. That is, 8 read from an image memory (not shown)
Bit image data is transferred to the shift register 10 one pixel at a time.
is input. Further, the image data that has passed through the delay circuit 1 is inputted pixel by pixel to the shift register 10 at the next stage.

Here, the delay time of the delay circuit 11 is one horizontal scanning time in the ITV camera. Here, as shown in FIG. 3, a coefficient table (RA
M) 15 at the same time as an address signal. In the table CRAM) 15, this input concentration data (
Access the data within the address indicated by the address signal) and output the access data as density data (16 bits).
Output as . However, a total of nine pieces of output density data set based on each coefficient parameter are input to the adder 13.

The adder 13 takes the sum S, and further adds the bias value BIAS (for example, 1020) to the sum S.
As a result, the upper 8 bits of the added data are subjected to a product-sum operation and output as normalized data, and this image data is written to a predetermined address in the image memory.

According to this embodiment as described above, the following effects are achieved. That is, in the past, product-sum operations and normalization operations were performed using multiplication, addition, and division, but in this embodiment, they can be realized only by accessing and adding data in table 15, so multiplication and division are not necessary. The processing speed can be increased accordingly. Therefore, spatial filtering processing can be performed at a speed equal to or faster than the input of a video signal from an ITV camera, for example, and this satisfies the real-time performance required when the image processing device is applied to a visual inspection device. It is. Further, costs can be reduced by eliminating the need for multipliers and dividers.

The present invention is not limited to the above embodiments, but may be modified and implemented as follows. That is, in the above embodiment, all 8 bits of the input density data are used as the address signal of the table 15, but the upper 5 bits are used as the address signal, and the other 3 bits are used as the control information (multiple weighting coefficient tables 15). It may also be used as a selection signal for selecting one from the following.

Further, in the above embodiment, the output density data of the table 15 is 16 bits, but it may be 8 bits. This is advantageous because the input data (corresponding to the multiplication value) to the adder 13 becomes 8 bits, which means that the circuit for the product-sum operation can be configured as an 8-bit processing system (conventionally, the weighting coefficient was up to 24). (then it becomes 12 bits). In this case, compared to the 16-bit method in the above embodiment, the accuracy of the calculation result will be lower due to the smaller number of bits, but considering the accuracy of the ITV camera, the accuracy of the A/D converter, etc., it is suitable for an image processing device. No major impact will occur.

Furthermore, the parameters of the spatial filter may be other than those shown in FIG. 8, and of course, in addition to the 3x3 matrix, a 5x5 matrix or a 7x7 matrix may be used. For the table 15, an element such as a ROM (read only memory) may be used in addition to the RAM.

In addition, the present invention can be implemented with various modifications without departing from the gist thereof.

[Effects of the Invention] As described in detail above, the image processing device according to the present invention includes a spatial filter processing unit that performs a sum-of-products operation and normalizes each pixel of a target image to be image-processed using a coefficient based on a spatial filter parameter. In the image processing device, the spatial filter processing section provides a conversion table that determines an output density value for each pixel density value of the input image based on coefficients based on the spatial filter parameters, and The configuration is such that an output pixel is determined for each pixel, and each of the output pixels is added.

According to this configuration, the product-sum operation and the normalization operation are realized using a conversion table without performing multiplication processing and division processing, thereby reducing the processing time of the spatial filter in the image processing device. can be provided.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of an image processing device according to the present invention, FIG. 2 is a diagram showing a coefficient table, FIG. 3 is a block diagram showing main parts of the embodiment, and FIG. 4 is a general diagram. 5 is a diagram for explaining the sum of products operation, FIG. 6 is a block diagram showing a conventional example, FIG. 7 is a diagram showing a specific example of the sum of products operation, and FIG. FIG. 8 is a diagram showing an example of parameters of the spatial filter. 10... Shift register, 11... Delay circuit, 13
... Adder, 15... Conversion table. Figure 3, X2,2 Input image data Figure 7, Figure 8, Y2,2 Output image data

Claims (1)

[Claims] In an image processing device that includes a spatial filter processing unit that performs a product-sum operation and normalizes each pixel of a target image to be image-processed using a coefficient based on a spatial filter parameter, the spatial filter processing unit A configuration in which a conversion table is provided in which an output density value is determined for each pixel density value of the input image based on a coefficient based on the input image, an output pixel is obtained for each pixel of the input image, and each of the output pixels is added. An image processing device characterized by: