JPS6379181A - Image processing device - Google Patents

Abstract

PURPOSE:To attain the Hough conversion with good accuracy at a high speed by subtracting a prescribed constant from the coordinate data of a point sequence to constitute an image, obtaining an arithmetic index value and incrementing it with a referring table based on said result. CONSTITUTION:By a host computer 1, a look-up table (LUT)6 is written. Next, an image memory 2 is swept and the coordinates (Xi, Yi) of a point sequence are stored into a RAM3. The loop is changed M times with an X0 from zero and a Y0 from -M/2 up to M/2. For the processing in the loop, an increment arithmetic part 7 and the RAM3 are set to zero, the RAM3 is swept in the sequence up to i=1-N and one loop processing is completed. An output (Xi, Yi) of the RAM3 is added with the output of registers 5a and 5b by adders 4a and 4b, Xi-X0 and Yi-Y0 are calculated and inputted to a LUT6. From the LUT6, an arithmetic index value Z corresponding to the inclination of a straight line to link a point (Xi, Yi) and a point (X0, Y0) is outputted, and by the increment arithmetic part 7, the contents shown by the value Z are incremented.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an image processing device that performs Hough transformation at high speed.

(Prior art) Recently, vanishing point application technology has been attracting attention, such as determining the direction of the camera by determining the vanishing point from a spatial image taken with a television camera. is necessary.

Conventionally, determination of straight lines by Hough transformation has been used in image processing, etc., but Hough 11 transformation is a process for extracting straight line components and circular arc components from a sequence of points. In this method, the coordinates of the point sequence are (X, .

yi), by finding the number of points that match the linear equation y = ax + b for all a and b, extracting the linear component and finding the linear equation, and also using the circle equation (x xo) 2
+ (y yo)” = the number of points that match r2,
All X. The equation of the circle is obtained by calculating for +3'o and r. In particular, in the case of a linear equation, it is necessary to take the size of a from -ω to +(1), but it is not possible to take it linearly, but when 0 = jan -'a, θ
It is ideal for accuracy to take linearity from -π to π.

However, in an image processing device that uses such a conventional 1101J g h conversion,
There is no dedicated processor for Ugh conversion, and
It takes time to calculate ', so the entire screen is
There was a problem in that conversion required a huge amount of calculation time.

(Object and structure of the invention) The present invention has been made in view of the above points, and [
The purpose is to shorten the processing time when processing images using Iough transformation. The calculation index value is calculated using a predetermined reference table, and the calculated calculation index value is incremented.

(Example) The present invention will be explained below based on the drawings.

FIG. 1 is a block diagram of an image processing apparatus according to the present invention.

First, to explain the configuration, the host computer (preferably a microprogram type) 1 is centered around the image memory 2,
RAM3, adders 4a, 4b, constant registers 5a, 5b
, an LUT (lookup table) 6, and an increment@ calculation section 7. The image memory stores two image data, and the RAM 3 is for storing point sequence address data, and can at least write X to the host computer 1 and output y to the adder 4a and y□.

If the image memory 2 is 256×256 (M=256), the data in the RAM 3 is X.

y can be 8 bits each, making the total 16 bits. Adder 4
a and 4b are the outputs of RAM3 (X, .

y+) and the outputs of registers 5a and 5b (-XQ, -yo)
Add. The contents of registers 5a and 5b are written by host computer 1. Outputs of adders 4a and 4b (
X = -X On yi-yo) is input to LUT6. LUT6 is composed of RAM and is connected to host computer 1.
After writing the contents, input X.

A calculation index value Z is output for Y. X and Y are each 8
-9 bits, Z about 8 bits (one image memory may be used) (RA of 64 to 256 bytes)
M). The output of the LUT 6 is input to the increment calculation section 7. The increment calculation unit 7 has a RAM of at least 256 bytes, and increments (+1)l, . . . and re-stores the contents of the address represented by the calculation index value Z.

The increment calculation unit 7 may be configured with a 256-tree counter.

In the image processing apparatus according to the present invention, after inputting an image, edges are emphasized and binarized. Then, tough conversion is performed on the "l" portion, that is, the edge portion.

FIG. 2 is a flowchart of Hough transformation for obtaining a straight line according to the present invention.

First, the host computer 1 writes to LUT6 (F, -1). In this case, the output (calculation index value) Z for the input (x, y) is when IYI≦IXI, that is, when the slope of the straight line a = the absolute value of Y/X or less than 1, Z =
When lyl>lXt, that is, when the absolute value of the slope a is 1 or more, Z=X/Y. In this way, -1<Z<1 can be obtained for -■<a<ω, and the error will be reduced even if the calculation is performed with 8-bit accuracy.

Next, the image memory 2 is swept and the coordinates (x i , y r ) of the point sequence forming the image are stored in the RAM 3 (F-
2). Let i=1 to N, where N is the number of points.

Varies depending on the image. Set XO to zero, change yo from -M/2 to M/2 (the image size is MXM), and perform the following processing (F-3). y.

corresponds to the intercept of the linear equation y=ax+b. ' Regarding the in-loop processing, first, the increment calculation section 7 is initialized and the value of each counter or the contents of the RAM is set to zero (F-4). Next t: RAM3 III! Sweep from i=1 to N (F-5). The outputs (xi, y+) of the RAM 3 are added to the outputs of the registers 5a and 5b by adders 4a and 4b, respectively, and X i-X On
'J 1-yO is calculated (F-6) and
, Y (F-7). From LUT6, the values corresponding to the inputs X=x, -xo+Y=y=-Yo, that is, a certain point (x t + y,) and the point (xo, yo)
A calculation index value Z corresponding to the slope of the straight line connecting the lines is output. Then, the increment calculation unit 7 increments the content of the address indicated by the calculation index value Z corresponding to this slope (F-8). In this way, RAM3 is
=1 to N, the frequency of the slope of the straight line connecting the point (X o + yO) and all the points (x i - y i ) remains in the increment calculation unit 7.

Finally, the contents αJ (j=1 to 2
56) and (j.

yo) position (where j corresponds to the slope a and Yo to the intercept) (F-9) and one loop ends.

This process is y. If the loop is repeated M times from =-M/2 to M/2, the linear (linear) Hough transformation of one screen is completed (F-10). FIG. 3(B) shows the coordinate system of an image obtained by Hough-transforming the different images shown in FIG. 3(A).

Next, FIG. 4 shows a flowchart of the process for determining the arc component according to the present invention.

Writing to LUT6 is performed using the slope of the straight line described above.
The value corresponding to the radius is X+V1. Sweep the image memory 2 and store the coordinates (X=, yi) of the point in the RAM 3 (P-2), change Xo from -M/2 to M/2, and yo also from -M/2 to M/2 Sweep the entire screen by varying it up to , and perform calculations within the loop at each point (P-3, P-4).
, the calculations in the loop are the same as in the case of a straight line; first, the increment calculation section 7 is initialized (P-5), and the RAM 3 is sequentially swept with i=1 to N (P-6). Next, (Xi, yi) of RAM output 3 is sent to register 5 by adders 4a and 4b.
The outputs of a and 5b are added (X.-x o), (y
r −y o) is calculated (P-7), and X
, Y (P-8), and the calculation index value Z output from LUT 6 is the point (X,) in this case.

Since the value corresponds to the distance between yi) and the point (xo, yo), it is incremented by the increment calculation unit 7 (
P-9), the content αJ (j=1 to 256) is the frequency of distance. Therefore, if the point (x o + y o ) is at the center of the circular arc, α will be a large value, and the value of j at that time corresponds to the size of the radius. Then, the contents αj of the increment calculation unit 7 are read out in order and α that is larger than a certain threshold value is read out.
The number of objects j is stored in the (xo, yo) position of the image memory (P-10) and one loop is completed. This process is

=-M/2 to M/2 and y. =-M/2 to M/2
If the entire screen is swept by performing the steps up to (P-11), the result will be an image that has a value at the center point of the arc.

(Effects of the Invention) As explained above, in the present invention, a predetermined constant is subtracted from the coordinate data of a sequence of points constituting an image, and a calculation index value is calculated using a predetermined reference table based on the subtraction result. Since the calculated calculation index value is incremented, it is possible to perform lough conversion with high precision at high speed.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the image processing apparatus according to the present invention, FIG. 2 is a flowchart showing the procedure for detecting linear components in the image processing device according to the present invention, and FIG. The coordinate system of the linear component obtained is shown, (a) is the image before conversion, (b) is the image after conversion, and FIG. 4 shows one song of the arc component detector in the image processing device according to the present invention. The flowchart, FIG. 5, shows the coordinate system of the arc component obtained by the present invention,
(a) is the original image, and (b) is the obtained image. l...Host computer, 2...Image memory, 3
-RAM, 4a, 4b-adder, 5a, 5b-register, 6...LUT (lookup table), 7...
Incremental operation unit patent applicant Hiroshi Suzuki, agent of Nissan Motor Co., Ltd., patent attorney Figure 1 Figure 3 (a) (b)
-X-'' Figure 5

Claims (1)

[Claims] An image memory that stores image data; a writable and readable memory that sequentially outputs coordinate data of a point sequence among the image data stored in the image memory; a subtraction means for subtracting a constant; a lookup table in which a predetermined reference table is stored in accordance with processing and outputs a calculation index value using the reference table based on the subtraction result from the subtraction means; and the lookup table. An image processing apparatus comprising: an incrementing means for incrementing the contents of an address indicated by an arithmetic index value output from an image processing apparatus.