JPH0555915B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an image transformation device that implements geometric transformation of an image using an image memory.

2. Description of the Related Art Special effects are given to images by performing geometric transformations of the images, such as rotation, enlargement, and reduction. In recent years, an image conversion device using an image memory has been proposed to realize this geometric transformation of an image.

Methods for realizing geometric deformation of an image using an image memory include methods based on write address control and methods based on read address control.

Write address control is based on information about which point in the output image each sample point of the dominant image corresponds to, and the address is controlled when writing to the image memory. It was done as it was done. On the other hand, read address control is based on information about which sample point in the input image each point of the output image corresponds to, and the address is controlled when reading from the image memory. It was designed to be carried out in a realistic manner.

Conventionally, in the write address control method, only one sample point of the image memory corresponds to one sample point of the input image, which has the disadvantage that, for example, the output image is intermittently generated in an enlarged part. Therefore, conventionally, this write address control has not been used except for simple reduction without interruption. On the other hand, with read address control, even when the image is enlarged, the same sample point in the image memory can be read out repeatedly, so there will be no interruption in the output image. However, this read address control does not provide information on which sample point in the input image each point of the output image corresponds to.
For example, calculations must be made by finding the inverse function of a predetermined transformation function, which is difficult for arbitrary geometric transformations. Therefore, conventionally, the type of geometrical deformation of an image that can be realized using read address control has been greatly limited.

In the end, it is often more intuitive to provide information (calculated using a predetermined transformation function) about which point in the output image each sample point of the input image corresponds to, and as mentioned above, the output image If it is possible to avoid intermittency, it would be more convenient to use write address control.

In view of this point, the present invention is designed to prevent intermittence from occurring in the output image when using write address control.

An embodiment of an image conversion device according to the present invention will be described below with reference to the drawings.

In Figure 1A, each sample point of the input image is indicated by a black circle ``●'', and on the X <sub>1</sub> and X <sub>2</sub> orthogonal coordinates, for example, its
It is placed at a position where the X <sub>1</sub> and X <sub>2</sub> components are integer values. In addition, in Figure 1B, each sample point of the output image memory is indicated by a white circle ``○'', and on the _Y1 , _Y2 orthogonal coordinates, it is located at a position corresponding to each sample point of the input image shown in Figure 1A. It was placed there. Although only a few sample points of the input image and sample points of the output image memory are shown in FIGS. 1A and 1B, in reality, a large number of sample points exist.

Here _, ^the conversion formula _{from the X 1} , X ₁ , _{X 2} ) ...(1) _{Y 2} = ² (X _{1 ,}
However, as shown in Figure 1B, the positions (y ₁ , y ₂ ) on the orthogonal coordinates of Y ₁ and Y ₂ (y ₁ = ¹ (x ₁ , x ₂ ), y ₂ = ² (x ₁ , x ₂ )
)
Suppose that it is converted to . In this case, according to the conventional write address control method, the position (y ₁ ,
y ₂ ) (y ₁ and y ₂ generally consist of an integer part and a decimal part, but for example, the position specified by the values y ₁ ′ and y ₂ ′ with the decimal part truncated) Only one sample point was selected, and the image information of one sample point (x ₁ , x ₂ ) of the input image was written into this one sample point. Therefore, when enlarging an image, for example, sample points where no image information is written are created in the output image memory between sample points where image information is written, resulting in gaps in the output image.

In this example, on the Y ₁ , Y ₂ orthogonal coordinates, the sample point (x ₁ , x ₂ ) of the conversion formula corresponds to a unit area centered on the sample point (x ₁ , x ₂ ) of the input image. A predetermined area S ((y ₁ ,
y ₂ ) as the center, and the area indicated by diagonal lines in Figure 1B is determined,
All the sample points of the output image memory included in this predetermined area S are selected as those corresponding to the sample points (x ₁ , x ₂ ) of the input image, and all the sample points of the selected output image memory are The image information of the sample point (x ₁ , x ₂ ) of the input image is written to the sample point.

On this Y ₁ , Y ₂ orthogonal coordinate, the position (y ₁ , y ₂ ) corresponding to the sample point (x ₁ , x ₂ ) of the input image
Four numerical values α ₁₁ , α ₁₂ , α ₂₁ , and α ₂₂ are used to determine the predetermined area S centered at . And these four numbers α ₁₁ , α ₁₂ , α ₂₁ , and α ₂₂ are
The above conversion formula, Y ₁ = ¹ (X ₁ , X ₂ ), Y ₂ = ²
Each of (X ₁ , X ₂ ) is determined using the partial differential value or difference value at the sample point (x ₁ , x ₂ ) of the input image. When using partial differential values, a ₁₁ = ∂ ¹ (X ₁ , X ₂ )/∂X ₁ | _{(X1,X2)=(x1,x2)} ×1/
2×α ……(3) a ₁₂ = ∂ ¹ (X ₁ , X ₂ )/∂X ₂ | _{(X1,X2)=(x1,x2)} ×1/
2×α ……(4) a ₂₁ = ∂ ² (X ₁ , X ₂ )/∂X ₁ | _{(X1,X2)=(x1,x2)} ×1/
2×α ……(5) a ₂₂ = ∂ ² (X ₁ , X ₂ )/∂X ₂ | _{(X1,X2)=(x1,x2)} ×1/
2×α...(6) When using the difference value, a ₁₁ = [ ¹ (x ₁ + 1, x ₂ ) - ¹ (x ₁ , x ₂ )] x 1/2
×α ……(7) a ₁₂ = [ ¹ (x ₁ , x ₂ +1) − ¹ (x ₁ , x ₂ )] × 1/2
×α ……(8) a ₂₁ = [ ² (x ₁ + 1, x ₂ ) − ² (x ₁ , x ₂ )] × 1/2
×α ……(9) a ₂₂ = [ ² (x ₁ , x ₂ + 1) − ² (x ₁ , x ₂ )] × 1/2
×α ……(10) is determined. Here, α is a coefficient for area correction, and α>1. That is, when an interval occurs between the predetermined areas S defined on the Y ₁ and Y ₂ orthogonal coordinates, this predetermined area is enlarged to eliminate the intermittency between the predetermined areas S. Incidentally, factors that cause the intermittency include approximation of the predetermined region S using a ₁₁ , a ₁₂ , a ₂₁ , and a ₂₂ and calculation errors.

At this time, _{Y 1 ,}
A predetermined area (parallelogram area) S as shown in FIG. 2B is defined on the _Y2 orthogonal coordinates. This parallelogram area S is located at the position (y ₁ , y ₂ ) corresponding to the sample point (x ₁ , x ₂ ) (y ₁ and y ₂ are the integer parts i ₁ and
i ₂ and fractional parts s ₁ and s ₂ . ) is defined by a vector whose Y ₁ component is a ₁₁ and whose Y ₂ component is a ₂₁ , and a vector whose Y ₁ component is a ₁₂ and whose Y ₂ component is a ₂₂ .

In the end, the sample points included within this parallelogram area S are selected as the sample points of the output image memory that should correspond to the sample points (x ₁ , x ₂ ) of the input image. Then, the image information at the sample point (x ₁ , x ₂ ) of the input image is written to the sample point of the selected output image memory, and the image conversion process at the sample point (x ₁ , x ₂ ) of the input image is performed. It will be done. By performing such processing over all sample points of the input image, image conversion processing is performed for the entire input image.

FIG. 3 is a block diagram showing the entire image conversion apparatus that performs such processing. In the same figure,
Reference numeral 1 indicates an input image memory, which is composed of, for example, RAM, and the image information of the sample point (x ₁ , x ₂ ) of the input image is transferred to the sample point (x ₁ , x ₂ ) of this input image memory 1. written. Further, in this FIG. 3, 2 is a processing device, 3 is an address generator, and 4 is a processing device.
is the output image memory.

Information on the conversion formulas Y ₁ = ¹ (X _{1 ,} X ₂ ), Y ₂ = ² (X ₁ , Address control and arithmetic processing for the image memories 1 and 4 are performed. Also,
This processing device 2 sequentially specifies sample points of an input image to be subjected to image conversion processing.

When a sample _point (x ₁ , x ₂ ) of an input image is specified, this processing device 2 immediately uses a conversion formula to determine the _position (y ₁ , y ₂ ) (y ₁ = ¹ (x ₁ , x ₂ ), y ₂ = ²
(x ₁ , x ₂ )) is calculated, and the above (3), (4), (5)
and (6), or according to (7), (8), (9) and (10),
The numbers a ₁₁ , a ₁₂ , a ₂₁ and a ₂₂ are calculated. In this case, a memory, for example, a ROM, is prepared in advance, in which the results of these calculations are written according to the conversion formula, and the processing device 2 reads y ₁ , y ₂ , a ₁₁ , a ₁₂ from this ROM. , a ₂₁ and a ₂₂ may be read. The position (y ₁ , y ₂ ) of the output image calculated by this processing device 2 and the numerical values a ₁₁ , a ₁₂ , a ₂₁ and
The information a ₂₂ is supplied to the address generator 3.

The address generator 3 is configured as shown in FIG. 4, for example. In the figure, 30 indicates an arithmetic circuit, and information of numerical values a ₁₁ , a ₁₂ , a ₂₁ and a ₂₂ is supplied to its input terminals 30a, 30b, 30c and 30d. In this arithmetic circuit 30, the numerical values a ₁₁ ,
Based on the information of a ₁₂ , a ₂₁ and a ₂₂ , various numerical values w ₁ , w ₂ , l ₁ , l ₂ , p ₁ are used to actually determine the parallelogram area S as shown in FIG. 2B, for example. and p ₂ are calculated. Among the numerical values that determine the parallelogram area S calculated by this arithmetic circuit 30, w ₁ , l ₁ and
p ₁ is supplied to the sample point calculation circuit 31A, and w ₂ ,
l ₂ and p ₂ are supplied to the sample point calculation circuit 31B.

Further, in this FIG. 4, 32 indicates an integer/decimal part separation circuit 32, and its terminals 32a and
2b is supplied with information on numerical values y ₁ (=i ₁ (integer part) + s ₁ (decimal part)) and y ₂ (=i ₂ (integer part) + s ₂ (decimal part)). The fractional parts s ₁ and s ₂ separated by this separation circuit 32 are supplied to a sample point calculation circuit 31A and also to a sample point calculation circuit 31B.

_The sample point calculation _circuit 31A calculates _a parallelogram area S on the orthogonal coordinates of _Y1 and _{Y2 ,} as _shown in FIG. _A
is determined. In the figure, white circles "○" are sample points of the output image memory 4. this area
S _A is s ₁ in the Y ₁ direction from the reference sample point (0, 0),
The center is set at a position (s ₁ , s ₂ ) separated by s ₂ in the Y ₂ direction. As a result, the sample point calculation circuit 31A calculates sample points included in this area _SA . Further, in the sample point calculation circuit 31B, the supplied numerical values w ₂ , l ₂ , p ₂ , s ₁
and s ₂ , Y ₁ , Y ₂ as shown in Figure 5
A region S _B is defined on the orthogonal coordinates. In the case of this region S _B , the center is also set at (s ₁ , s ₂ ). As a result, this sample point calculation circuit 31B
Then, sample points included in this area S _B are calculated. These sample point calculation circuits 31A and 3
Information on the sample points calculated in 1B is supplied to the common sample point calculation circuit 33. This calculation circuit 33 calculates common sample points included in the areas S _A and S _B , that is, sample points that should be included in the parallelogram area S in FIG. 2B. In this case, as is clear from the previous explanation, the calculated sample point is the reference sample point (0,0)
The position is shown relative to

This common sample point information is supplied to the address generation circuit 34. This address generation circuit 34
are supplied with information on the integer parts i ₁ and i ₂ of y ₁ and y ₂ separated by the separation circuit 32 . This address generation circuit 34 converts the reference sample point (0, 0) to (i ₁ , i ₂ ) by being supplied with the information of i ₁ and i ₂
At the same time, the above-mentioned common sample points are specified on the output image memory 4. Based on this, the address generation circuit 34 specifies a common sample point on the output image memory 4. That is, an address signal ADo specifying a sample point that may be included within the parallelogram area S is generated and taken out to the output terminal 35.

Address signal generated by address generator 3
ADo is supplied via the processing device 2 to the output image memory 4 along with the control signal Sco.

Further, the processing device 2 supplies the input image memory 1 with an address signal ADi and a control signal Sci specifying the sample point (x ₁ , x ₂ ). Then, the image information _ID of the sample point (x ₁ , x ₂ ) of the input image written to the sample point (x ₁ , x ₂ ) is read out and sent to the output image memory 4 via the processing device 2. Writing is performed by being supplied to the sample point specified by the address signal ADo mentioned above.

In this way, according to the image conversion device shown in FIG. 3, the processing device 2 specifies the sample points of the input image, performs the above-described processing on each of them, and performs the above-mentioned processing over the entire input image. Image conversion processing is performed.

As described above, according to the image conversion apparatus according to the present invention, arbitrary geometric deformation is relatively easy because it is based on write address control.
Moreover, the sample point of the input image is wrapped in a predetermined area (area according to the rate of change of the image) determined on the output image memory according to the sample point and the conversion formula. Compared to the conventional method that simply corresponds one sample point of the input image to one sample point of the output image memory, it can be used even when the rate of change in the image is large, for example when enlarging the image. ,
Furthermore, even when a transformation function other than linear transformation is given, sample points with no image information are not generated in the output image memory, and there is no risk of intermittency in the output image.

Next, another embodiment of the image conversion apparatus according to the present invention will be described with reference to FIG. 6 and subsequent figures.

In this example, each sample point of the input image (indicated by a black circle "●") is arranged as shown in FIG. 6A.
_A predetermined area _{S 0} is defined on _the X ₁ , According to the predetermined region _S0 , the predetermined region S~ is plotted on the _Y1Y2 orthogonal coordinates where each sample point (indicated by a white circle " _○ ") of the output image memory is arranged as shown in FIG. 6B. determined. This predetermined area S~ is centered at the position (y ₁ , y ₂ ) corresponding to the sample point (x ₁ , x ₂ ) of the input image. Thereafter, each of the sample points of the output image memory included in the predetermined area S~ is associated with a selected pair of sample points of the input image included in the predetermined area _S0 .

The conversion formula is Y ₁ = ¹ (X _{1 ,} X _{2 )} ...(11) Y ₂ = ² (X _{1 ,} S ₀
is defined in an _m × _n rectangular shape as shown in FIG _. , _y2 )
(Parallelograms defined by four numbers a _{~11 ,} a~ ₁₂ , _a ~ ₂₁ , and a~ ₂₂ , centering on y1 = ¹ (X1, _X2 ) ^, _y2 = 2 ( _X1 , _X2 ) Shape area.Four numbers a~ ₁₁ , a~ _1
2 , a~ ₂₁ and a~ ₂₂ are calculated using partial differential values or difference values at sample points (x ₁ , x ₂ ) of the input image for each of the conversion formulas shown in equations (11) and (12) above. It can be decided. When using partial differential values, a~ ₁₁ = ∂ ¹ (X ₁ , X ₂ )/∂X ₁ | _{(X1,X2)=(X1,X2)} ×m
/2×α……(13) _a〜12 =∂ ¹ (X ₁ , X ₂ )/∂X ₂ ｜ _{(X1,X2)=(X1,X2)} ×n
/2×α……(14) _a〜21 =∂ ² (X ₁ , X ₂ )/∂X ₁ | _{(X1,X2)=(X1,X2)} ×m
/2×α……(15) _a〜22 =∂ ² (X ₁ , X ₂ )/∂X ₂ | _{(X1,X2)=(X1,X2)} ×n
/2×α……(16) When the difference value is used, a~ ₁₁ = [ ¹ (x ₁ + m, X ₂ ) − ¹ (x ₁ , X ₂ )] × 1
/2×α……(17) a~ ₁₂ = [ ¹ (x ₁ , X ₂ + n) − ¹ (x ₁ , X ₂ )] × 1
/2×α……(18) a~ ₂₁ = [ ² (x ₁ + m, X ₂ ) − ² (x ₁ , X ₂ )]×1
/2×α……(19) a~ ₂₂ = [ ² (x ₁ , X ₂ + n) − ² (x ₁ , X ₂ )] × 1
/2×α...(20)

Here, α is a coefficient for correcting the size of the parallelogram area, and a>1. In other words, Y _{1 , Y 2 according to the predetermined area S 0 defined on the X 1 X 2 orthogonal coordinates}
If an interval occurs between the parallelogram areas S~ defined on the orthogonal coordinates, this parallelogram area S~ is expanded to eliminate the interval between the parallelogram areas S~. The factors that cause the intermittence include approximation of the predetermined area wave S using a~ ₁₁ , a~ ₁₂ , a~ ₂₁ , a~ ₂₂ , calculation errors, etc.

In this way, a parallelogram area S~ is defined on the _Y1 , _Y2 orthogonal coordinates corresponding to the predetermined area _S0 defined on the _X1 , _X2 orthogonal coordinates, and within the predetermined area _S0 . Among the sample points of the output image memory, the sample points contained within the parallelogram region S~ are selected as corresponding to the sample points of the input image contained therein.

However, each of the sample points of the output image memory included in this parallelogram area S~ is a predetermined area.
It is unknown which sample point of the input image included in S ₀ corresponds to. Therefore, next, sample points of the input image included in the predetermined area _S0 corresponding to each of the sample points of the output image memory included in the parallelogram area S~ are determined. That is, it is determined which sample point in the output image memory each sample point of the input image included in the predetermined area S ₀ should correspond to.

Now, the point (y ₁₀ , y ₂₀ ) in the parallelogram area S that corresponds to the point (x ₁₀ , x ₂₀ ) in the predetermined area S ₀ is (y ₁₀ y ₂₀ ) = (y ₁ y ₂ ) +(a ₁₁ ′, a _{12 ′} a ₂₁ ′, a ₂₂ ′) (X _{10 −X 1} In this formula, a ₁₁ ′=2/ma~ ₁₁ , a ₁₂ '=2/na~ ₁₂ , a ₂₁
'=2/ma _{~21 , a22} '=2/na~ ₂₂ .

In this example, from this, each of the sample points of the output image memory included in the parallelogram area S~ corresponds to which sample point among the sample points of the input image included in the predetermined area _S0 . The inverse function of the linear approximation formula shown in equation (21) above is used to determine whether the The inverse function of this linear approximation formula is (x ₁₀ x ₂₀ ) = (x ₁ x ₂ ) + (b ₁₁ b ₁₂ b ₂₁ b ₂₂ ) (y ₁₀ −y ₁ y ₂₀ −y ₂ ) ... (22 ). In this equation (22), (b ₁₁ b ₁₂ b ₂₁ b ₂₂ ) is (a _1
1 ′ a ₁₂ ′ a ₂₁ ′ a ₂₂ ′).

Here, among the sample points of the output image memory included in the parallelogram area S~, the sample point ( _y11 ,
y ₂₁ ), the point (x ₁₁ , x ₂₁ ), (x ₁₁ x ₂₁ ) = (x ₁ x ₂ ) + (b ₁₁ b ₁₂ b ₂₁ b ₂₂ ) (y ₁₁ −y ₁ y ₂₁ − y ₂ ) ...(23) is determined within the predetermined area S ₀ and the sample point (y ₁₁ ,
As sample points of the input image to correspond to y ₂₁ ),
For example, a sample point (x _{11 ′} , x _{21 ′} ) within a predetermined area S ₀ close to the point (x ₁₁ , x ₂₁ ) is determined. Regarding other sample points of the output image memory included in the parallelogram area S~, the sample points of the input image included in the corresponding predetermined area _S0 are determined in the same manner.

In this way, the sample points of the output image memory corresponding to the sample points of the input image included in the predetermined area S ₀ defined on the input image are determined. In this case, the sample points of the output image memory corresponding to the sample points of the input image are the sample points of the output image memory corresponding to the sample points of the input image, that is, the sample points of the output image memory corresponding to the sample points of the input image, in the above-described embodiment. This is approximately the same as the sample point of the output image memory included in the predetermined area S defined on the output image memory. Eventually, the image information at the sample point of the corresponding input image is written to the sample point of this output image memory, and
The image conversion process at _S0 ends.

Note that the predetermined areas S ₀ as described above are sequentially taken on the input image, and image conversion processing is performed on the entire input image.

FIG. 8 is a block diagram showing the entire image conversion apparatus that performs such processing. In FIG. 8, parts corresponding to those in FIG. 3 are designated by the same reference numerals.

The processing device 2 includes the conversion formulas Y ₁ = ¹ (X ₁ , X ₂ ), Y ₂ = ² (X ₁ , X ₂ ) shown in equations (11) and (12) above.
information and a predetermined area defined on the X ₁ and X ₂ orthogonal coordinates
Information on numerical values m and n that determine the size of S ₀ is input in advance, and address control and arithmetic processing for the image memories 1 and 4 are performed. Furthermore, the processing device 2 sequentially specifies sample points of the input image that are to be the center of a predetermined area S ₀ sequentially taken on the X ₁ and X ₂ orthogonal coordinates.

When a sample point (x _{1 ,} x ₂ ) of an input image is specified, this processing device 2 uses _a conversion formula to convert (Y ₁ , position ( _{y 1 ,} y ₂ ) (y ₁ = ¹
(x ₁ , x _{2 )} , y ₂ = ² (x ₁ , ), (19) and (20), the numerical values a~ ₁₁ , a~ ₁₂ , a~ ₂₁ and a~ ₂₂ are calculated.

The information on the position (y ₁ , y ₂ ) on the output image and the numerical values a~ ₁₁ , a~ ₁₂ , a~ ₂₁ and a~ ₂₂ calculated by this processing device 2 is supplied to the address generator 3 .

This address generator 3 is constructed as shown in FIG. In this address generator 3,
Based on the information of the numerical values a~ ₁₁ , a~ ₁₂ , a~ ₂₁ and a~ ₂₂ ,
Various numerical values w~ ₁ , w~ ₂ , l~ 1 , l~ ₂ , p _{~ 1 and} p~ ₂ that determine the parallelogram area S~ as shown in FIG. 7B
is calculated. Then, from these values, areas S~ _A and S~ _B are determined on the output image (on _Y1 , _Y2 orthogonal coordinates) as shown in Figure 10, and these areas S~ _A and S~ _B
Common sample points included in , that is, sample points to be included in the parallelogram area S~ in FIG. 7B are calculated. Then, these sample points become the sample points (y ₁₁ ,
y ₂₁ ), the address generator 3 generates an address signal AD~o specifying the sample point (y ₁₁ , y ₂₁ ) on the output image memory 4.
is generated. Then, this address signal AD~o
is supplied via the processing device 2 to the output image memory 4 together with the control signals Sc to o, and is also supplied to the address generator 5 .

The address generator 5 is configured as shown in FIG. 9, for example. In the figure, 50 indicates an inverse matrix calculation circuit, and terminals 50a, 50b, 50c, and 50d are supplied with numerical values a~ ₁₁ , a~ ₁₂ ,
Information a~ ₂₁ and a~ ₂₂ is supplied. The arithmetic circuit 50 calculates the inverse matrix (b ₁₁ b ₁₂ b ₂₁ b ₂₂ ) of (2/ma ₁₁ 2/na ₁₂ 2/ma ₂₁ 2/na ₂₂ ). Numerical values b ₁₁ , b ₁₂ ,
The information of b ₂₁ and b ₂₂ is supplied to the address generation circuit 51. Terminal 51 of this address generation circuit 51
The above-described address signal AD~o (having information on the sample point (y ₁₁ , y ₂₁ ) within the parallelogram area S~) is supplied to the terminals 51b, 51c, and 51a.
d and 51e, numbers x ₁ , x ₂ , y ₁ are set by the processing device.
and y ₂ information is provided. In this address generation circuit 51, the sample point (y ₁₁ ,
The sample point of the input image (x ₁₁ , y ₂₁ ) corresponding to
x ₂₁ ) is calculated, and image information corresponding to the sample point (x ₁₁ , x ₂₁ ) of the input image is written to the output terminal 51f. An address signal AD~i designating a sample point in the input image memory 1 is taken out.

This address signal AD~i obtained by the address generator 5 is supplied via the processing device 2 to the input image memory 1 together with the control signal Sc~i.

When the address signal AD~i is supplied to the input image memory 1, the image information I~ _D of the sample point of the input image is read out from the sample point of the input image memory 1, and is processed via the processing device 2. It is supplied and written to the sample point (y ₁₁ , y ₂₁ ) designated by address AD~o in the output image memory 4 .

From this address AD~o, the sample point (x ₁₁ , x ₂₁ ) of the input image is calculated, and the image information of the sample point (x ₁₁ , x ₂₁ ) of the input image is read out from the input image memory 1 and used for output. The process of writing to the sample points (y ₁₁ , y ₂₁ ) in the image memory 4 is repeated as many times as the number of sample points (y ₁₁ , y ₂₁ ) included in the parallelogram area S~.

In this way, the sample points (x ₁ ,
Image conversion processing is performed on a predetermined area S ₀ centered on x ₂ ).

In this manner, according to the image conversion apparatus shown in FIG _.
are sequentially determined and the above-described processing is performed on each of them, so that image conversion processing is performed on the entire input image.

As described above, this embodiment has the same effects as the above-mentioned embodiment, and it is not necessary to define a predetermined area on the output image for each sample point of the input image. It is convenient because the amount of processing can be reduced.

According to the above embodiment, the image information obtained from the input image is once written into the input image memory, but the image information may be directly supplied from the input image to the output image memory.

[Brief explanation of the drawing]

1, 2 and 5 are diagrams for explaining an embodiment of the image conversion device according to the present invention, FIG. 3 is a block diagram showing one embodiment, and FIGS. 4 and 9 are FIG. 6, FIG. 7, and FIG. 10 are diagrams for explaining other embodiments, and FIG. 8 is a block diagram showing another embodiment. . 1 is an input image memory, 2 is a processing device, 3 and 5 are address generators, and 4 is an output image memory.

Claims (1)

[Claims] 1. An image transformation device configured to transform an input image based on a set transformation function to form an output image, wherein a plurality of inputs each including a plurality of sample points on the input image are provided. a region forming means for defining a small region and forming a plurality of output small regions corresponding to the plurality of input small regions based on a transformation function of a representative point of each input small region; and a conversion means for assigning the image information signal of each sample point in the input small area to each sample point in the output small area based on the inverse function for each output small area. An image conversion device characterized by: 2. The region forming means is configured to form a plurality of output small regions corresponding to the plurality of input small regions, respectively, based on partial differential values at each representative point of a transformation function of a representative point of each input small region. An image conversion device according to claim 1, characterized in that: 3. The region forming means is configured to form a plurality of output small regions corresponding to the plurality of input small regions, respectively, based on a difference value at each representative point of a conversion function of a representative point of each input small region. An image conversion device according to claim 1, characterized in that: