JPS62287378A - Nearby picture image arithmetic unit - Google Patents

Abstract

PURPOSE:To process simultaneously process plural different-density picture elements at the same time by providing a nearby picture element register, a bit selector, and a bit computing element. CONSTITUTION:Data on plural nearby picture elements [respective (m) bits of (n) of nearby picture elements] read out of an image memory 1 are latched at a time by the nearby picture element register 2 consisting of (n) registers. Outputs A1-An of the register 2 are connected by the bit selector 3 to the bit computing element 4 successively in units of the same bit section of each picture element with an instruction from a program sequencer 5. Groups of the same digits are connected to the bit computing element 4 in the order of the MSB, next digit, further next digit of each register 2. For example, when (n)=9 and intensity information on an image consists of 16 bits, the selection of the bit section of the bit 6 inputs following data to the bit computing element 4.

Description

Detailed Description of the Invention 3. Detailed Description of the Invention (Field of Industrial Application) The present invention relates to a neighborhood image calculation device that performs calculation processing between adjacent pixels.

(Prior Art and Problems) Conventional neighborhood image arithmetic units for grayscale images carry out processing for each pixel between the pixels as shown in Figure 3 as a method for performing calculations regarding grayscale information in the vicinity. A method using a combination of binary operations on intensity information is known. FIG. 3 shows an example of a circuit that performs a combination of binary operations on the intensity information of eight pixels.

With this method, if it is desired to obtain information regarding the relationship between the intensity information of a large number of pixels constituting a neighborhood, it is necessary to change the processing procedure in various ways.

For example, it is difficult to perform processing such as determining the position of a pixel with the highest intensity among a large number of input pixels using the same arithmetic unit.

SUMMARY OF THE INVENTION An object of the present invention is to provide a neighborhood value calculation device that can collectively process a plurality of gray pixels input at the same time.

(Means for Solving the Problem) In order to achieve the above object, a neighborhood image calculation device according to the present invention provides a neighborhood image calculation device for calculating data of a plurality of pixels having multivalued intensity information obtained from an image. In the 112 arithmetic unit, a neighboring pixel register is provided corresponding to a plurality of pixels included in a processing unit and launches data for each pixel, and a bit selector sequentially reads out bit sections from the neighboring pixel register. , a bit arithmetic unit that performs a batch operation on each piece of data included in a read bit section, determines arithmetic processing conditions for the next bit section according to the result, and repeats the operation.

(Example) The present invention will be described in more detail below with reference to the drawings and the like.

FIG. 1 is a block diagram showing an embodiment of a neighborhood image calculation device according to the present invention.

The data of adjacent pixels to be subjected to calculation from the image memory 1 are latched in a nearby pixel register (latch) 2 all at once.

Data of a plurality of neighboring pixels (each m bits of n neighboring pixels) read out from the image memory 1 is latched at once in a neighboring pixel register 2 consisting of n registers.

The bit selector 3 sequentially connects the outputs A1 to An of the register 2 to the arithmetic unit 4 for the same bit cross section of each pixel according to a command from the microprogram sequencer 5.

Groups of the same digits, such as the MSB of each register 2, the next digit, and the next digit, are connected to the bit arithmetic unit 4.

For example, if n=9 and the intensity information of the image is composed of 16 bits, when the bit cross section of bit 6 is selected with the bit selector, the following data is stored in the bit operator 4.
is input.

That is, bit 6 of A, bit 6 of A2...
.

A total of 9 bits, bit 6 of A9, are bit operations'? 54 inputs.

FIG. 2 is a block diagram showing an embodiment of the bit arithmetic unit 4. As shown in FIG.

The bit operation unit 4 includes a PLA (programmable logic array) 40. Pattern generation ROM41. Lookup table 42. Arithmetic operator 43. Register file 44
．． It is composed of a latch 45.

PLA 40 is a ROM with programmable addresses. The PLA 40 performs logical operations (for example, AND and OR for each bit) on the bit pattern generated by the pattern generation ROM 41 and the input bit section.

For a bit string manually input in parallel, it is possible to calculate the number of points where the bit in the data is 1, its digit T, bit pattern matching, etc.

The lookup table 42 calculates the number of 1's in the bit string, detects the bit position, and performs pattern matching.

The resulting bit string or numerical result is fed back to the pattern generator 41.

Further, the pattern generation output of the pattern generator 41 is also controlled by the output of the register file 44.

As a specific calculation, an example will be shown below in which the position of the pixel having the maximum value among 3×3 neighboring pixels is detected.

In FIG. 1, n=9, and all pixels forming a 3×3 neighborhood are latched into the neighborhood pixel register 2.

Thereafter, the bit selector 3 sequentially moves from the bit cross section 15 toward the bit cross section O, and while lowering the position of the bit cross section, the following calculations and judgments are performed by the bit arithmetic unit 4.

As an initial state, the latch 45 (see FIG. 2) is cleared, and the bit string 111111111 is input from the pattern generator 41 to the input B of the PLA 40.

Also, the registers in the register file 44 are cleared.

Step 1: Perform AND logic between the bit string input to the input A of the PLA 40 and the data 111111111 input to the input.

Step 2: The number of points whose output bit is 1 is calculated by the look amplifier table 42, and compared with the numerical value 1 by the ALU 43.

Step 3: If the sum is 1 or the bit cross section has been selected down to 0 bits, step 6 jumps.

*The total is 1, which means that only one pixel has data 1 and the others are O, so the maximum ii! This means that element j has been determined.

Step 4: Input the output from the latch 45 of the bit string output from the PLA 40 to the pattern generation ROM 41, generate a bit pattern string with focus 1 only for the bit position that was 1, and input it to input B of the PLA 40. .

That is, from now on, only the data of the pit position that was 1 will be considered.

Step 5: Select the bit cross-section lowered by one bit by the bit selector 3 and input it to the input A of the PLA 40.

Then step 1 jump.

Step 6: Calculate the pit position where 1 is located using the lookup table 42 and output it to the status output.

As described above, calculations on intensity information between pixels can be processed all at once.

(Effects of the Invention) As explained in detail above, the neighborhood image calculation device according to the present invention performs processing on data of a plurality of pixels having multivalued intensity information obtained from an image. a neighboring pixel register that is provided corresponding to a plurality of pixels included in a unit and latches data for each pixel; a bit selector that sequentially reads data from the neighboring pixel register for each bit section; It consists of a bit arithmetic unit that performs a batch operation on each piece of data included in a bit section, determines arithmetic processing conditions for the next bit section according to the results, and repeats the operation.

Therefore, according to the present invention, the following effects can be obtained.

○Compared to conventional arithmetic units, the same arithmetic unit can perform general-purpose calculations that determine the relationship between density information between pixels.

Regarding the above, the number of steps is significantly smaller than that of conventional arithmetic units, so it is faster.

0 The circuit configuration is simpler than conventional arithmetic units, and the number of parts can be reduced.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of a neighborhood calculation device according to the present invention. FIG. 2 is a block diagram showing a configuration example of a bit arithmetic unit of the neighborhood arithmetic unit. FIG. 3 is a block diagram showing an embodiment of a conventional neighborhood calculation device. 1... Image 1 image memory 2... Neighboring pixel register (launch) 3... Bit selector 4... Bit arithmetic unit 41... Pattern generation ROM 42... Lookup table 43... Calculation Device 44...Register file 5...Microprogram sequencer Patent applicant Hamamatsu Photonics Co., Ltd. Agent Patent attorney Jusai Inoro 1 Figure L-Funtora 4% 9v upper-J Figure 2 λCas↑-Vs Output

Claims (2)

[Claims] (1) In a neighborhood image calculation device that calculates data of multiple pixels having multivalued intensity information obtained from an image,
a neighboring pixel register that is provided corresponding to a plurality of pixels included in a processing unit and latches data for each pixel; a bit selector that sequentially reads data from the neighboring pixel register for each bit section; A neighborhood image arithmetic device comprising a bit arithmetic unit that performs a batch operation on each piece of data included in a bit cross section, determines arithmetic processing conditions for the next bit cross section according to the result, and repeats the operation. (2) The neighborhood image calculation device according to claim 1, wherein the calculation is performed by calculating the maximum value of a plurality of pixels having multivalued intensity information obtained from the image.