JPS6354245B2 - - Google Patents

Description

[Detailed Description of the Invention] [Technical Field of the Invention] This invention relates to an image processing device,
More specifically, the present invention relates to an n-dimensional digital filtering device that performs variable digital filtering on input data at high speed and in real time to enhance images.

[Technical background of the invention]

In recent years, two-dimensional data such as medical X-ray images have been stored in digital memories, allowing various filtering techniques, and image enhancement, smoothing, etc., to be carried out quantitatively. For this reason, taking medical X-ray images as an example, the X-ray exposure dose has been reduced compared to conventional techniques, and high-quality images can now be obtained in a short time.

[Problems with background technology]

Incidentally, the above-mentioned filtering is performed by calculations performed by a computer, and in many cases, filtering operations are performed using binary numbers using a digital memory. Therefore, when there is a large number of data such as images,
Filtering in real time is difficult, and, taking medical image processing as an example, it has been an obstacle for doctors to quickly obtain diagnostic data.

[Purpose of the invention]

The present invention has been made based on the above-mentioned circumstances, and provides an n-dimensional digital filtering device that can quantitatively perform digital filtering in a real-time variable manner and can quickly process effective images. The purpose is to provide

[Summary of the invention]

The outline of this invention for achieving the above object is as follows:
An n-dimensional digital filtering device that performs digital filtering on n-dimensional data includes a data storage means for storing n-dimensional data as an analog quantity, and a real-time variable digital filter coefficient for the n-dimensional data as an analog quantity. and a convolution means that sequentially performs convolution of n-dimensional data and digital filter coefficients in analog quantities and outputs the results as a digital signal in real time. That is.

[Embodiments of the invention]

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of a two-dimensional digital filtering device which is an embodiment of the present invention.

In FIG. 1, the two-dimensional digital filtering device includes an operation console 1 for giving various instructions from the outside, a control device 2 that operates according to instructions from the operation console 1 to control the entire device, and a control device 2 for controlling the entire device. An input device 3 for inputting data, a filtering device 4 for storing image data as an analog quantity such as electric charge and performing filtering, and a filter design device 5 or filter bank 6 for setting filter coefficients for the filtering device. and an output device 7 for displaying the output of the filtering device 4 on a tape play or outputting it to another storage device.

Next, the filtering device 4, which plays a central role in the two-dimensional digital filtering device, will be explained in detail with reference to FIGS. 2, 3, and 4. In FIG. 2, the filtering device 4
An image data storage device 8 stores two-dimensional image data from the input device 3 in a buffer, for example a charge transfer element, and sequentially transfers the image data in multiple rows.
and input the output of the image data storage device 8.
a convolution device 9 that performs dimensional convolution; and a controller 10 that controls operations of the image data storage device 8 and the convolution device 9 based on control signals from the control device 2.
and a filter coefficient storage device 11 that stores the filter coefficients transferred from the filter design device 5 or the filter bank 6 as a charge amount, etc., and outputs them for convolution in the convolution device 9. It is completed.

FIG. 3 shows the image data storage device 8, convolution device 9 and filter coefficient storage device 11.
2 is a block diagram showing the specific configuration of the

In FIG. 3, the image data storage device 8
consists of a D/A converter 12 and, for example, k image buffers 13-1, 13-2, . . . 13-k. The D/A converter 12 is necessary only when the input data is a digital quantity, and is not necessary when the input data is an analog quantity.
The k image buffers 13-1 to 13-k are
It consists of elements that input k rows of image data and each store one row of image data. Here, as an example, k image buffers are used so that the number of lines is the same as the size k of the filter support. 1st row image buffer 13-
The details of 1 will be explained with reference to FIG.
It consists of charge storage elements 13-1-1 to 13-1-k and image buffers 13-2 to 13-k in other rows.
k also has a similar configuration.

The filter coefficient storage device 11 is composed of k filter coefficient buffers 14-1 to 14-k that store filter coefficients transferred from the filter bank 6 or the filter design device 5. Here, as an example, a digital filter with filter coefficients k×k, that is, a square digital filter is used. The details of the filter coefficient buffer 14-1 in the first row will be explained with reference to FIG.
1 is k charge storage elements 14-1-1 to 14-
1-k, and the filter coefficient buffers 14-2 to 14-k in other rows have similar configurations.
Note that it is also possible to utilize the positive and negative four symmetries or other symmetries of the digital filter coefficients, so k×k filter coefficient buffers are not necessarily required. Further, the coefficients stored in the filter coefficient buffers 14-1 to 14-k may be the coefficients transferred from the filter bank 6 or the filter design device 5 and may be retained as they are during filtering. Alternatively, the filter coefficients may be newly changed at the same time as the real-time filtering process. That is, during real-time high-speed digital filtering, the filter coefficients are variable.

The convolution device 9 stores image data from the image buffers 13-1 to 13-k and the filter coefficient buffers 14-1 to 14-k. k analog multipliers 15-1 to 15-k that input filter coefficients for each row and multiply them by analog quantities; an analog adder 16 that adds these multipliers; and an analog adder 16 that adds these multipliers; D
It consists of an A/D converter 17 that converts and outputs the data.

The details of the analog multiplier 15-1 in the first row among the analog multipliers 15-1 to 15-k are described in the fourth section.
To explain with reference to the figure, the analog multiplier 15-
1 is the filter coefficient buffer 14- in the first row.
Each charge storage element 14-1-1 to 14- in 1
1-k to input the filter coefficients from k
It consists of multipliers 15-1-1 to 15-1-k. Analog multipliers 15-2 to 15 in other rows
-k also has a similar configuration.

The operation of the two-dimensional digital filtering device configured as above will be explained.

Two-dimensional image data as a digital amount is output from the input device 3 based on control signals from the control device 2 in response to operations on the console 1, and is sequentially input to the filtering device 4. In the filtering device 4, first, the image data storage device 8 stores the two-dimensional image data in an analog quantity. That is, D/A
The converter 12 converts the two-dimensional image data into analog quantities, and stores k rows of two-dimensional image data as charge quantities in k image buffers 13-1 to 13-k for each row. For example, in the image buffer 13-1 in the first row, the two-dimensional image data in the first row is used as the density value of the pixel in k charge storage elements 13.
-1-1 to 13-1-k.

On the other hand, the filter coefficients set in advance in the filter design device 5 or the filter bank 6 are stored in the filter coefficient storage device 11 in the filtering device 4 under the control of the control device 2.
It is stored as a ×k digital filter.
Note that, as described above, this filter coefficient may be newly changed at the same time as the filtering process. The k×k filter coefficients are k filter coefficient buffers 14-1 to 14.
-k, and for example, in the filter coefficient buffer 14-1 in the first row, k charge storage elements 14-
1-1 to 14-1-k, the filter coefficients are stored as charges.

The two-dimensional image data stored in the image data storage device 8 and the filter coefficients stored in the filter coefficient storage device 11 are subjected to two-dimensional convolution in a convolution device 9. First, in k rows of analog multipliers 15-1 to 15-k in the convolution device 9,
Multiplication of the two-dimensional image data and the filter coefficient is performed for each row. For example, in the analog multiplier 15-1 in the first row, the two-dimensional image data stored in the image buffer 13-1 in the first row and the one
The filter coefficient stored in the filter coefficient buffer 14-1 of the row is input and multiplied.
More specifically, in the n multipliers 15-1-1 to 15-1-k in the analog multiplier 15-1, the charge storage elements 14-1-1 to 14-1-k in the filter coefficient buffer 14-1 are is multiplied by the two-dimensional image data from the charge storage elements 13-1-1 to 13-1-k in the image buffer 13-1.
Multiplication is performed in the same manner in each of the other rows. Each of these multiplication results is input to an adder 16, and a two-dimensional convolution result is obtained as an output of this adder 16. The output of the adder 16 is then converted into a digital signal by an A/D converter 17, and can be displayed as an image on the output device 7, for example.

As explained above, since digital filtering can be performed using analog quantities, real-time filtering processing is possible even when the amount of data is large, such as image data.

It goes without saying that the present invention is not limited to the embodiments described above, and includes various modifications within the scope of the gist of the invention. The data to be subjected to filtering is not limited to two-dimensional image data, but can be applied to various types of data, and each member shown in the above embodiment may be another member having the same function.

〔Effect of the invention〕

As explained above, according to the present invention, n-dimensional data can be digitally filtered using an analog quantity and then converted to a digital value.
In addition, since the variation of filter coefficients is not hindered during filtering execution, it is possible to provide an n-dimensional digital filtering device that can perform high-speed variable filtering in real time.

Therefore, even when filtering a large amount of data such as image data, it is possible to perform filtering at high speed.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a two-dimensional digital filtering device which is an embodiment of the present invention, FIG. 2 is a block diagram of the filtering device shown in FIG. 1, and FIG. 3 shows the specific configuration of the filtering device. Block diagram showing,
FIG. 4 is a block diagram for explaining data processing for one line in the filtering device. 4... Filtering device, 8... Image data storage device, 9... Convolution device, 11...
...Filter coefficient storage device, 12...D/A converter, 13-1 to 13-k...Image buffer, 14
-1 to 14-k...Filter coefficient buffer, 15
-1 to 15-k...analog multiplier, 16...adder, 17...A/D converter.

Claims (1)

[Claims] 1. In an n-dimensional digital filtering device that performs digital filtering on n-dimensional data, there is a data storage means for storing n-dimensional data as an analog quantity, and a real-time variable digital filter coefficient for n-dimensional data as an analog quantity. filter coefficient storage means for storing n
An n-dimensional digital filtering device comprising: convolution means for sequentially convoluting dimensional data and digital filter coefficients in analog quantities and outputting the results as a digital signal in real time.