JPH03282977A - Image processor - Google Patents

Abstract

PURPOSE:To selectively execute the convolution(CV) processing of data with high operation necessity by reading out the data of points to be computed and CV coefficient data in accordance with order indicated by the output of a sequence counter and executing CV processing. CONSTITUTION:A sequence counter 1 generates an order signal 2 indicating the order of operation and applies the signal 2 to a ROM 3 and a coefficient memory 9 as a reading address signal. Convolution(CV) coefficient data 11 to be the output of the memory 9 are read out in accordance with the order of operation and inputted to a multiplier 12. On the other hand, points to be computed are read out from the ROM 3 in the order indicated by the signal 2 and inputted to an image memory 6 as a reading address 4. Only computed point data 8 are read out from the memory 6 and the multiplier 8 executes the CV processing of the data 8, 11 and outputs a processed result.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an image processing device that performs two-dimensional convolution processing on image data.

(Prior Art) An image processing device that performs two-dimensional convolution processing on image data is not a device that handles computer graphic images, but a device that handles natural images such as photographic images.
Specifically, it is used, for example, when performing edge enhancement processing to remove blurred outlines of photographic images.

In this type of image processing device, image data is stored in an image memory in blocks each consisting of N×N calculation target points.
Further, the convolution coefficient data is stored in a coefficient memory, and the convolution process is performed using the image data of all calculation target points and all the convolution coefficient data.

(Problem to be solved by the invention) In the conventional image processing device described above, N
Since convolution processing is performed on all points in the XN matrix, there is a problem in that the processing takes a long time.

The present invention has been made in view of such conventional problems, and its purpose is to provide an image processing device that can selectively perform convolution processing on image data at points where calculation is highly necessary. There is a particular thing.

(Means for Solving the Problems) In order to achieve the above object, an image processing apparatus of the present invention has the following configuration.

That is, the image processing device of the present invention comprises: an image memory in which image data subjected to two-dimensional convolution processing is segmented into a plurality of calculation target point data and stored; and convolution coefficient data corresponding to the calculation target point data. an image processing device comprising: a coefficient memory in which is stored; a sequence counter that outputs a signal indicating the order of calculations and also outputs the signal to the coefficient memory as a read address signal; A calculation target point memory in which a calculation target point of data is stored in advance, and in which a corresponding calculation target point is read out in response to an output signal of the sequence counter and is outputted to the image memory as a read address signal. It is characterized by having the following features.

(Function) Next, the function of the image processing apparatus of the present invention configured as described above will be explained.

In the present invention, corresponding calculation target point data and convolution coefficient data are read out according to the order indicated by the output of the sequence counter, and subjected to convolution processing.

In other words, the convolution process is not performed on all of the calculation target point data in the image memory, but only the data of points that are highly necessary for calculation are selectively subjected to the convolution process.

As a result, processing time can be significantly reduced.

(Example) Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows an image processing device according to an embodiment of the present invention.
In FIG. 1, this image processing device has a sequence counter 1 and a ROM 3 serving as a calculation target point memory placed before the image memory 6 and coefficient memory 9 in the conventional device.
Furthermore, a control signal 5 is added.

In the image memory 6, image data 7 is stored for each calculation target block consisting of NXN calculation target points. Also,
The coefficient memory 9 stores convolution coefficient data lO corresponding to each calculation target point data. From these, calculation target point data 8 and convolution coefficient data 11 are read out as will be described later. The calculation target point data 8 and the convolution coefficient data 11 are provided by the multiplier 1.
2, the multiplication result 13 is sent to the accumulator 14.
An accumulation operation is performed at , and a processing result I5 is output.

Hereinafter, parts related to the present invention will be explained. The sequence counter 1 is a circuit that generates a signal (sequence signal) 2 indicating the order of operations.The generated sequence signal 2 is applied to the ROM 3 and the coefficient memory 9 as a read address signal. As a result, the convolution coefficient data 11 that is the output of the coefficient memory 9 is read out in accordance with the order of calculations.

On the other hand, the ROM 3 stores the calculation target points necessary for the calculation among the calculation target points of the image data, and the calculation target points stored at the address indicated by the order signal 2 are read out and are read out as the read address signal 4. The image is applied to the image memory 6. As a result, only the calculation target point data 8 necessary for the calculation in the calculation target block is read out from the image memory 6 and output. A specific example will be explained with reference to FIG.

FIG. 2 (a> shows one calculation target block in the image memory 6. This calculation target block consists of 7 x 7 calculation target points whose X and Y coordinates are both O to 6. The circle-marked part in the figure indicates the calculation target point necessary for the calculation, and the unmarked part in the figure indicates that it is not the target point.And,
○The numbers in the egg (1 to 21> indicate the order of calculation, and in this calculation target block, the first calculation target point (X, Y)
This shows that the data of = (3', 3) is used, and in the 21st time, the data of the calculation target point (X, Y) = (3, 6) is used. That is, in this calculation target block, only data of 21 calculation target points out of 49 calculation target points are read out and subjected to calculation.

FIG. 2(b) shows the contents of the ROM 3 for reading data of the block to be calculated shown in FIG. 2(a). As shown in the figure, in this ROM 3, each address from 0 to 20 corresponds to the order of operations (1 to 21), which is the content of the order signal 2.
Corresponding to pair 1, the Y-coordinate value and X-coordinate value that specify the calculation target point corresponding to the calculation order are set, and the end flag is set to 0. For example, corresponding to the calculation order "1" At address "0", Y=3. X=3 is set. As a result, in the image memory 6, data at the calculation target point (X, Y)=(3, 3) is read out in the calculation order "1". In addition, convolution coefficient data 11
has a one-to-one correspondence with the calculation target point data 8, so
In the coefficient memory 9, like the image memory 6, only data necessary for calculation is read out.

Also, the end flag indicates that the order of the calculation in the block to be calculated is the highest! & Later, only the address “20” corresponding to the sequence number “21” is set to ``I'', and all others are set to ``0゛''. Therefore, the contents of address "20" are read out. At this time, the end flag 1'' is read out and given as a control signal 5 to the sequence counter 1 and the accumulator 14.

In the sequence counter 1, the input of the control signal 5 causes the sequence counter to become 1.
Knowing that the reading of the calculation target points for the block has been completed, the process proceeds to the output operation of the order signal 2 for the next block. On the other hand, the accumulator I4 learns from the input of the control signal 5 that the calculation for 11072 minutes has ended, and outputs the processing result 15.

(Effects of the Invention) As explained above, according to the image processing device of the present invention,
Since the calculation target point data and convolution coefficient data can be read out according to the order of calculation, it is possible to selectively perform convolution processing only on the data of points that require calculation, significantly reducing processing time. This has the effect of making it possible.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of the configuration of an image processing apparatus according to an embodiment of the present invention, and FIG. 2 is an explanatory diagram of the relationship between a ROM and an image memory. 1...Sequence counter, 3...
ROM, 6... Image memory, 7...
Image data, 9...Coefficient memory, 12...
...multiplier, 14...accumulator.

Claims (1)

[Claims] An image memory in which image data subjected to two-dimensional convolution processing is divided into a plurality of calculation target point data and stored; and a coefficient memory in which convolution coefficient data corresponding to the calculation target point data is stored. In an image processing apparatus; a counter that outputs a signal indicating the order of calculations, and a sequence counter that also outputs the signal as a read address signal to the coefficient memory; and a sequence counter in which calculation target points of the image data are stored in advance. a calculation target point memory, which receives an output signal from the sequence counter, reads out a corresponding calculation target point, and outputs it to the image memory as a read address signal;
An image processing device comprising: