JPS5886658A - Compensating device for geometric distortion of picture - Google Patents

Abstract

PURPOSE:To perform precise, easy and high-speed compensating calculation, by dividing a picture into partial areas, and compensating geometric distortion of each partial area according to a linearly approximated compensation function. CONSTITUTION:A control unit 3 sets the center coordinates (x0, y0) of one of divided partial areas in an input picture memory 1 in a coordinate register 6. According to this data and coefficient data set in a coefficient register 8, a coefficient calculating circuit 9 calculates and sets the coefficient of the linearly approximated primary coordinate conversion of a compensation function in a coefficient register 7. Then, X.Y coordinate calculating circuits 4 and 5 input data on the center coordinates (x0, y0) set in the register 6 and the primary coordinate conversion coefficient set in the register 7, and also inputs a variable from the control unit 3 to calculate position coordinates (X, y) having geometric distortion compensated, specifying an address of an output picture memory 2.

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field of the Invention The present invention relates to a highly practical geometric distortion correction device that can easily correct geometric distortion of an image at high speed.

<Technical Background of the Invention> Images captured and input by a television camera (ITV), scanner images from an artificial satellite, etc. generally have nonlinear geometric distortion. However, by converting such an image into a digital image and using a computer system,
When processing i*, use some means to process the above lm1j
It is necessary to correct the geometric distortion of I. The correction function for this geometric distortion is usually given approximately as a two-variable higher-order equation. For example, the coordinates of an image with geometric distortion are (X
#y), if the coordinates of this image after geometric distortion correction are (X, Y), then the coordinates (X, Y) are the coordinates (x*y
) can be expressed as (x s y )→(x, y). ]! If we show each coordinate component of the above mapping using functions f and l, then X=f (X @
y ) Y - p (x @ y ), and this function is generally well approximated by a polynomial of degree 2 or higher.
y"Y-b@+blx+b@y+bHx"tenb4xy+
b@yl''. Therefore, using this approximation function, it is possible to effectively correct the geometric distortion of the good image described above.

<Problems with conventional technology> However, performing correction calculations for image geometric distortion using the above-mentioned correction functions is complicated and complicated. As a result, it is not possible to connect the calculation processing, and it is also necessary to provide a special calculation circuit to execute the calculation sensitivity layer exclusively. It is difficult to perform geometric distortion correction, and it is impractical and difficult to perform.

<Object of the Invention> The present invention has been made in consideration of the above circumstances, and its object is to easily and quickly perform correction calculations for non-linear 'geometric distortion of an image. An object of the present invention is to provide a highly practical image geometric distortion correction device.

<Summary of the Invention> The present invention divides an image having geometric gold into a plurality of partial regions, and performs geometric distortion correction for each of these partial regions according to a linearly approximated correction function. . This makes it possible to precisely correct the geometric distortion of the image as a whole, and to perform correction calculations easily and at high speed for individual partial areas by linear coordinate transformation.

In other words, the quadratic or higher order 4 given as a geometric distortion correction function
Taking advantage of the fact that even the following function can be partially approximated to a linear equation, the image can be divided into multiple subregions, and each of these subregions can be linearly transformed to perform geometric distortion correction. This is what I did.

<Embodiment of the Invention> An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows a schematic configuration of the apparatus of the embodiment, in which 1 is an input image memory that stores an input image having geometric distortion inputted by a television camera (ITV) or the like;
Reference numeral 2 denotes an output image memory that stores a corrected image obtained by correcting the geometric distortion of the input image, that is, an output image. The input image storage 1 is configured to receive address designation (xey) from the control unit S, store it at the corresponding address, and generate the data of the nine input images. The output image memory 2 has the control knee.

X, coordinate calculation - M4 and Yji that operate under SO control! Address specification (X, Y) 1 by standard calculation circuit 5
The image data read out from the image memory 1 is written into the input address. These X-Y coordinate calculation circuit 4elFi, coordinate position information tied to the coordinate register 6, and coefficient register 1
The coordinates (X, Y) of the output image memory 2 are mapped to the coordinates (xey) of the input image i1 according to the correction function information given from the control unit 3 according to the linear coordinate transformation coefficient set in . are calculated respectively. It should be noted that the coordinate position information stored in the coordinate register 6 indicates nine individual partial areas into which the input image is divided into a plurality of areas, as will be described later, and consists of, for example, address information representing each partial area. The linear coordinate transformation coefficients set in the coefficient register 1 are calculated by the coefficient calculation circuit 9 from the coordinate position information from the coordinate register 6 and the coefficients of the approximate expression of the geometric distortion correction function given to the coefficient register 8. They respectively correspond to the determined partial areas. First, the above-mentioned nine correction functions for the geometric distortion of the input image are provided to the control unit 3. Under the timing control by this control unit 3, the respective parts described above in 6 operate in conjunction with each other, and the input 1ti
It is constructed so that the good image data stored in the i*memory 1 is coordinate-changed and written into the output image memory 2, where its geometric distortion is corrected.

Now, the input image stored in the input image memory 1 is, for example, data of MXN pixels, Km as shown in FIG.
It is divided into a plurality of partial regions each consisting of X n pixels. For each of these partial area lines and the entire input image, data is set as the 111th line in the M pixel direction and the 1st line in the N pixel direction.

J>K] Each one is specified. Then, the six points of the partial area are indicated as the partial area <i, J>O center position If (x., y.).

Regarding the partial area <i, j>0 center ff1lli(!・,y・) divided in this way, if the correction formula for the geometric distortion of the image given by control 22 and IK is expanded to the first-order term, we get the following: X straw!・+α−131・+β Totomo case. However, it is given as: X・: @×U−1)+− y・=n This Taylor-expanded nine-corrected formula is a linear formula with respect to the variables α and 7.

Therefore, as mentioned above, the correction function is bivariate a2
When given as the following equation, μ um-al+a4x+2a"7 y,,-h@+2bBx+h4y p,,w-b@10b4x+2bm7, so the expanded approximation is X"f(x* sy*)+( at +2a Lux* +a
4y)Q+(&1+AyaX・+2agy・)β Y−P(x@,y@)+(bs+2bs!
@ + ha ye) α+ (bB+b4x@+2km
1m) becomes β. It becomes a linear equation with respect to the variables α and β. Therefore, the image of the partial region <s, j> is subjected to coordinate transformation using the constant terms /(z・shi・), F(X・,c) in the above equation and the coefficients of the variables α, /. Do it, ζ
In this way, it becomes possible to hydraulically adjust the geometry ★ of the image of the partial region <S*j>. However, the above-mentioned variables α, βO and the range of numbers are 11≦α≦−1≦β≦−2222 with the positional accuracy of the representative point as the center (X·, C). Then, the geometric distortion correction of the partial region <i, j> by such -dimensional coordinate transformation is performed as follows: 1≦1≦−, 1≦j≦
By repeatedly increasing the depth over a range of 7 mm, geometric distortion correction is performed over the entire input image.

Therefore, according to the present apparatus having the configuration shown in FIG. information and an approximate expression of a correction function for the geometric distortion of the image obtained by another means or the like is input. After obtaining such information, the controller 22 and 1 first set the coefficients of the approximate expression of the correction function in the coefficient register 1.

This coefficient setting to the coefficient register 8 is stored in the input image memory 1 and is performed once before geometric distortion correction of a good image. Then, using these set pieces of information, the geometric distortion correction process of the image is executed as follows.

That is, first, the control unit 3 inputs ii! Iiga
One of the divided partial areas on memory 1 is specified, and the center coordinates (X·.) of this specified partial area are specified.

1.) is set in the coordinate register 6. Based on the data of the center coordinates (X・, y・) set in the coordinate register 6 and the coefficient data set in the coefficient register 8, the coefficient calculation circuit 9 calculates the − Calculate the coefficients of the approximated linear coordinate transformation and set the coefficient values obtained thereby in coefficient register 7-2
ing. This coefficient value calculation is performed each time the partial area to be corrected is updated.

In the X-Y coordinate calculation circuit 4.5, the data of friend center coordinates (X・, y・) set in the coordinate register and the linear coordinate transformation coefficient set in the coefficient register 7 are input, and 2. Variables (α, β
), and the positional loci 1i (X, Y) corrected for geometric distortion are calculated according to the above-mentioned linear approximation and nine correction formulas. This calculated position coordinate (X, Y
) specifies the address of the output image memory 2.

Then, the image data read from the address (x*y) of the input mI memory 1 is written to the address (X, Y) of the output image memory 2. Such coordinate calculations are sequentially performed over the entire partial area using the above (α, β) as variables, and the geometric distortion correction for the image of the partial area is completed. Thereafter, the next partial area is designated and similar processing is executed, and thereafter geometric distortion correction is repeatedly performed over the partial area of the entire screen. Therefore, correction processing for all partial regions is completed, and simple and accurate geometric distortion correction by linear coordinate transformation is completed.

<Effects of the Invention> As described above, according to the present invention, it has nonlinear geometric distortion,
Even if the image has a high-order correction function and is difficult to correct, it is possible to accurately correct geometric distortion by performing linear coordinate transformation for each of the nine divided regions and using simple means. Moreover, since only linear coordinate transformation is performed according to a −th order approximated correction function, calculation processing is simple and high-speed processing is possible. Furthermore, the configuration of the calculation processing circuit is simple and easy to implement using a conventional address calculation circuit. Furthermore, since the amount of data for company calculation processing is significantly reduced, it is possible to perform corrections with high accuracy, which has great practical effects.

<Other Embodiments of the Invention> Note that the present invention is not limited to the above-mentioned embodiments. , as mentioned above, it is possible to obtain a first-order approximation by dividing it into partial regions and capturing the above correction function on the circle. Therefore, in a case like Jin, To-yT:, is also a geometric distortion due to -dimensional coordinate transformation. Correction is possible. Therefore, it is possible to easily perform normalization of 1,000 degrees, and the effect of correcting nonlinear geometric distortion is extremely large.In short, the present invention can be implemented with various modifications without departing from the gist thereof.

[Brief explanation of drawings]

FIG. 1 is a schematic configuration diagram of an apparatus showing an embodiment of the present invention, and FIG. 2 is a schematic diagram showing the concept of divided partial regions of an image. ' 1... Input image memory, 2... Output image memory, 3
...Control unit, 4...X coordinate calculation circuit, 5...
・Y coordinate calculation circuit, 6...Coordinate reno star, 1...Coefficient register, 1...Coefficient register, 9...Coefficient calculation circuit・

Claims (2)

[Claims] (1) means for providing a correction function for geometric distortion of an input image; and means for dividing the input image into a plurality of partial regions;
means for respectively obtaining a linear coordinate transformation correction function for the input image of each of the partial areas according to the positional relationship of each of these partial areas with respect to the input iii* whole and the correction function; and means for calculating the transformed coordinate positions of the input image of each of the partial regions. (2) The -order coordinate correction function obtains a first-order approximation correction function by using the Taylor family of nine correction III numbers given for geometric distortion, and calculates the coefficients of this -order approximation correction function for the partial region and the entire input image. An image geometric distortion correcting device according to claim 1, wherein the correction is determined according to the positional relationship of the image.