JPH0827835B2 - Parallel projection information generation method - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a projection information parallel generation method for calculating information for controlling image enlargement / reduction processing by a computer at high speed.

Scaling of the prior art image original _{A (a 1, a 2 ......} , a m) for the results _{B (b 1, b 2 ......} b n) processing to obtain, i.e., the index of the original pixel This is an operation of calculating an index j of a result pixel that should correspond to a certain i and substituting the value of ai for bj.
(However, ai and bj are the values of each pixel of the original image and the result.) In this scaling process, there is a method of obtaining the result pixel bj by mapping the original pixel ai into a straight line. This will be described below.

FIG. 11 shows a case where the original pixel m bits are reduced to the result pixel n bits (m ≧ n). Original pixel A in the x-axis direction
(A ₁ , a _2, ..., _Am ) are arranged in the y-axis direction as a result pixel B.
A horizontal grid line h is assumed at the boundary between the result pixels with the arrangement of (b ₁ , b _2, ..., B _n ). It is assumed that the index j of h has the same value as the result pixel immediately above. A straight line L in the drawing is a straight line having a reduction ratio n / m as an inclination and an initial value R ₀ of a value (hereinafter, referred to as a residual) representing a quantization error as a y intercept. The correspondence from the original picture A to the result B is this straight line L
It can be obtained from the map.

The intersection of the perpendicular line passing through the center of the original pixel ai and the straight line L is Pi. When the result pixel penetrated by the horizontal line passing through the point Pi is bj, the index i of the original pixel corresponds to the index j of the result pixel, and the value of ai is transferred to bj. ai
When it is known in advance that it corresponds to bj, the method of obtaining the result pixel corresponding to the adjacent pixel ai + 1 of ai can be simplified. That is, ai + 1 corresponds to either bj or bj + 1, which can be determined by ascertaining whether or not there is a horizontal grid line hj + 1 between the point and Pi + 1.

In this way, it is possible to obtain the correspondence relationship between the original pixel and the result pixel by the method of mapping on the straight line L. Vector T (t ₁ , t ₂ , ……, t _m ) with the same elements
(Hereinafter referred to as a projection vector). The element ti of the projection vector T has information about which result pixel the original pixel ai corresponds to, and is determined by the following procedure.

If it is known that ai and bj correspond to each other, ai + 1 corresponds to either bj or bj + 1. For example, if the horizontal grid line hj + 1 is interposed between the points Pi and Pi + 1, then ai + 1 corresponds to bj + 1. When the horizontal grid line hj + 1 does not exist between the points Pi and Pi + 1, ai + 1 corresponds to bj. In the former case, the value of ti + 1 is 1, and in the latter case, the value of ti + 1 is 0. That is, when the value of ti + 1 becomes 1, the point
This is a case where pi + 1 exists above the horizontal grid line hj + 1 or on the horizontal grid line hj + 1, and when the value of ti + 1 becomes 0, it is the other case. When the value of t ₁ is 1, it means that the point p ₁ is above the horizontal grid line h ₁ or on the horizontal grid line h ₁ , and when the value of t ₁ is 0 Is the case. For example, a case where 7 bits of original pixels are reduced to 3 bits of result pixels will be described with reference to FIG. As described above, the original pixel A is set on each of the x-axis and the y-axis.
(A ₁ , a ₂ , ..., A ₇ ) Assume the arrangement of the result pixels B (b ₁ , b ₂ , b ₃ ). In addition, horizontal grid lines h ₁ , h ₂ , h ₃ between the result pixels
Draw. The index j of the horizontal line h has the same value as the result pixel immediately above, as shown in FIG.
The slope of the straight line L in the figure is 3/7. This slope 3/7 is a value with the original pixel number as the denominator and the resulting pixel number as the numerator. or,
R ₀ , which is the y-intercept of the straight line L, is the initial value of the value representing the quantization error described above, that is, the residual initial value.

The intersections of the perpendicular line passing through the center of the original pixel and the straight line L are p ₁ , p ₂ , ..., P ₇ from the left. The value of t ₁ is determined by whether or not the point p ₁ is located above the horizontal grid line h ₁ . In FIG. 12, the point p ₁ is located above the horizontal grid line h ₁ .
Therefore t ₁ = 1.

Next, determine the value of t ₂ . When obtaining the second and subsequent values of the projection vector T, as described above, it is necessary to have information as to which result pixel the original pixel immediately before the projection vector T corresponds to. In FIG. 12, the point p ₁ is located above the horizontal grid line h ₁ , and it is known that the original pixel a ₄ corresponds to the result pixel b ₁ . From this, the value of t ₂ is 1, the other becomes zero in the case of when the point p ₂ is either located on top of the horizontal grid lines h _2, or on the horizontal grid lines h _2. Therefore t ₂ = 0.

Next, consider the value of t ₃ . Since the original pixel a ₂ corresponds to the result pixel b ₁ , it may be determined whether or not the point p ₃ is located above the horizontal grid line h ₂ . In Fig. 12, the point p ₃ is
It is located above the horizontal grid line h ₂ . Therefore t ₃ = 1.

When t ₇ is sequentially obtained as described above, the projection vector T becomes T = (1010010) (1).

Let yi be the y-coordinate value of point Pi, and the residual at this point
Ri is Ri = yi-i (2). (However, i is the height of the horizontal line hi immediately below the point Pi. In addition, R _{5 in} FIG. 12 represents the residual.) The projection vector and the residual are calculated as described above. In the case of enlargement, the original pixel is made to correspond to B (b ₁ , b ₂ , ..., B _n ) and the resulting pixel is made to correspond to A (a ₁ , a ₂ , ..., _Am ), and the projection vector is , Are the same.

The projection vector is calculated by quantizing the intersection point pi of the perpendicular line drawn from the center of the original pixel ai and the straight line L in correspondence with the grid point, which is calculated by the conventional DDA (digital differential analysis) algorithm. It was obtained by sequential processing using an arithmetic unit. The actual calculation process will be described below.

When the reduction ratio is n / m, the length between the grid lines is m and the residual R
The value obtained by adding n to i is compared with m. When the calculation result Ri + n is larger than or equal to m, the i + 1-th element of the projection vector T, ti + 1
Is 1, and the value of the next residual Ri + 1 is Ri + n−m. If it is smaller, the value of ti + 1 is set to 0 and the next residual Ri +
The value of 1 is Ri + n. (For details of this algorithm, see K. Kawakami & S. Shimazaki (K.Ka
wakami and S. Simazaki), "A Special Purpose LSI Processor Equipped the Dayda Algorithm for Image Transformation (A Special Purpose LSI Proc
essor Using The DDA Algrithm For Image Transformat
ion) ”IEEE, Com Ark
Arc), 11th, 1984. ) Next, FIG. 13 shows this flow. 80 to 82 are processes for setting initial values of the element counter and the residual register. 83 to 86 are processes for calculating one element of the projection vector and the residual. 87 is a process of incrementing the original pixel counter and the projection vector element pointer by 1. Reference numeral 88 is a process of determining whether or not m elements of the projection vector have been calculated. If not, the process returns to 82 and the above process is repeated.

As described above, conventionally, the process can be performed by obtaining the elements and residuals of the projection vector for each one.

Problems to be Solved by the Invention However, in the conventional method as described above, since the above-described projection vector generation is sequentially performed by a program, a large number of arithmetic circuits are required, and it takes time and enlargement / reduction is required. It was extremely difficult to increase the processing speed.

In view of the above-mentioned conventional technique, the present invention aims to speed up the scaling process.

Means for Solving the Problems The present invention represents a correspondence relationship between a first pixel column composed of J pixels and a second pixel column composed of K pixels via a mapping to a straight line. A process of obtaining a projection vector composed of L bits (L is a large number of J and K) P bits (L> P) at a time in parallel is performed by a reduction ratio between the first pixel row and the second pixel row. A projection matrix generation process for obtaining a projection matrix of P rows and P columns from n, m for determining n / m <1 and a given quantization residual, and the next projection matrix generation following the projection matrix generation process. Residual candidate generation processing for obtaining P + 1 candidates of quantized residuals used in processing from n, m and the quantized residual used in the projection matrix generation processing, and projection obtained by the projection matrix generation processing. The projection vector determination process for determining the elements of the P projection vectors from the matrix and the residual candidate generation process are performed. Therefore, the P + 1 residual candidates generated and the quantized residual used in the next projection matrix generation process are determined from the elements in the Pth row of the projection matrix of P rows and P columns obtained by the projection matrix generation process. The residual determination process is performed Q times (Q ≧ 1).

Operation The present invention is to increase the speed by simultaneously calculating the elements of the P projection vectors when the residual R is given by the above configuration.

Example One example of the present invention will be described below.

In order to simplify the explanation, a description will be given of the case where the projection information in the reduction is set to 4 bits in parallel.

In FIG. 7, an array of original pixels A (a ₁ ,
a ₂ , a ₃ , a ₄ ), and the resulting pixel array B (b ₁ ,
b ₂ , b ₃ , b ₄ ), and a vertical grid line g passing through the center of each original pixel and a horizontal grid line h passing through the boundary between the result pixels. The index i of the vertical grid line g is the same as the value of the original pixel penetrated by gi, and the index j of the horizontal grid line h is
It is the same as the index of the result pixel just above hj. The grid lines are equally spaced in both horizontal and vertical directions,
Let l be its length. Slope n / m (m ≧ n), initial residual R ₀
Let Pi be the point of intersection of the straight line L with y as the y-intercept and the vertical cross line gi. Y-coordinate y ₁ of the point p ₁ is a value obtained by adding n / ml to the initial value R ₀ of the residual. Also, since the difference between yi and yi + 1 is n / m, Is.

Next, when the intersection of the vertical grid line gi penetrating the center of ai and the horizontal grid line hj is Qij, the height Qij is Qij = j × l = hj- (4). However, hj represents the height of hj.

Next, the matrix C (hereinafter,
Call the projection matrix).

From the above equations (3) and (4), Can be written. That is, the element Cij represents whether the point pi is above or below the grid point Qij. That is, if the value of the element Cij is 1, the point pi is located on or above the horizontal grid line hj, and if the value of the element Cij is 0, the point pi is located below the horizontal grid line hj. are doing.

Next, in the case of reduction, since n / m1, the projection matrix C satisfies the following condition.

Cij = 0; i <j Therefore, the projection matrix C is Becomes

The projection matrix C 6 in FIG. ₆ is Is. From the first row of this projection matrix C ₆ , the point P ₁ is the grid point Q
_It can be seen that it is above ₁₁ and below Q ₁₂ . Second
The row shows that point P ₂ is above grid point Q ₂₁ and below grid point Q ₂₂ . By the third line, the point P ₃ is the grid point Q.
You can see that it is above ₃₂ and below Q ₃₃ .
By the fourth line, point P ₄ is above grid point Q ₄₂ , and Q ₄₃
You can see that it is in the lower rank of.

As described above, the i-th row of the projection matrix C has information about which grid point the projection point pi of the original pixel ai on the straight line L exists. From this projection matrix C, the projection vector T
Is possible. This method will be described below.

The first element t ₁ of the projection vector is equal to the value of C ₁₁ . Because, if the value of C ₁₁ is 1, the point P ₁ is located above the horizontal grid line h ₁ . Therefore, the value of the element t ₁ of the projection vector is 1. If the value of C ₁₁ is 0, the point P ₁ is located below the horizontal grid line h ₁ . Therefore, the value of the element t ₁ of the projection vector is 0.

Next, the method of obtaining ti + 1 differs depending on whether the value of ti is 0 or 1.

(I) When ti = 1 (shown in FIG. 9) Of the elements having a value of 1 in the i-th row of the projection matrix, the maximum index j is defined as Ci and jmax. That is, the point pi
Is above the grid points Qi, jmax, and the grid points Qi, jmax + 1
Is in the lower part of. Therefore, the value of ti + 1 which is the i + 1th element of the projection vector is equal to the value of Ci + 1 and jmax + 1. That is, for Ci + 1 and jmax + 1, the point pi + 1 is the grid point Qi +
It is an element to judge whether it is higher than 1, jmax + 1. If the value of Ci + 1, jmax + 1 is 1, point pi + 1 and point
This means that the horizontal grid line h jmax + 1 exists between pi, and the value of the element ti + 1 of the projection vector is 1. When the values of Ci + 1 and jmax + 1 are 0, the point pi + 1 is below the horizontal grid line h jmax + 1. That is, point pi + 1 and point
Since it means that there is no horizontal grid line between pi, the value of the element ti + 1 of the projection vector is zero.

(Ii) When ti = 0 (shown in FIG. 10) Among the elements having a value of 0 in the i-th row of the projection matrix, the minimum value of the index j is jmin. That is, the point pi is the grid point
It is lower than Qi and jmin and higher than the lattice point Qi and jmin-1. Therefore, the i + 1th element ti + 1 of the projection vector
Is equal to the value of Ci + 1, jmin. That is, Ci + 1, jmin
Is an element for determining whether or not the point pi + 1 is above the lattice points Qi + 1, jmin. When the value of Ci + 1, jmin is 1, the horizontal lattice line h jmin is between the points pi + 1 and pi. Means that there exists, and the projection vector element ti + 1
Has a value of 1. If Ci + 1 and jmin are 0, point pi +
1 is located below the horizontal grid line h jmin. That is, it means that there is no horizontal grid line between the point pi + 1 and the point pi.
The value of +1 is 0.

As described above, from the projection matrix C, the projection vector T
Can be uniquely determined.

Now, it is also possible to obtain the projection matrix C from the projection vector T.

For example, a method of obtaining the projection matrix C _{7 in} the case of a 4-bit projection vector T ₇ (1010) will be described below with reference to FIG. 7.

First, as the first processing, the element of the first row of the projection matrix is obtained.

The point P ₁ is higher than the lattice point Q ₁₁ , and the lattice point Q ₁₂ ,
Lower in the Q _13, Q _14. Therefore, (C ₁₁ , C ₁₂ , C ₁₃ , C ₁₄ ) = (1, 0, 0, 0) − (8) Next, as the second processing, the element of the second row of the projection matrix is obtained.

Point P ₂ is higher than grid point Q ₂₁ , and grid point Q _{22 is}
Lower in the Q _23, Q _24. Therefore, (C ₂₁ , C ₂₂ , C ₂₃ , C ₂₄ ) = (1, 0, 0, 0) − (9) Then, as the third processing, the element of the third row of the projection matrix is obtained.

Point P ₃ is higher than grid points Q ₃₁ and Q ₃₂ , and
It is lower than Q ₃₃ and Q ₃₄ . Therefore, (C ₃₁ , C ₃₂ , C ₃₃ , C ₃₄ ) = (1, 1, 0, 0) − (10) Then, as the fourth processing, the element in the fourth row of the projection matrix is obtained.

Point P ₄ is above grid points Q ₄₁ and Q ₄₂ , and grid point Q ₄₃ and Q _42
It is under ₄₄ . Therefore, (C ₄₁ , C ₄₂ , C ₄₃ , C ₄₄ ) = (1,1,0,0) − (11) From the above (8), (9), (10), (11), the estimation matrix
C ₆ is Becomes

As described above, the projection vector T can be obtained from the projection matrix C. On the contrary, the projection matrix C can be obtained from the projection vector T. That is, the projection vector T and the projection matrix C have a one-to-one correspondence.

Next, from the projection matrix C _{6 in} FIG. 7, the projection vector T
_{A case of} obtaining ₆ (t ₁ , t ₂ , t ₃ , t ₄ ) will be described.

As mentioned above, the value of t ₁ is equal to element C ₁₁ . Therefore, t ₁ = C ₁₁ = 1- (13).

As the value of t ₂ , the case of (i) above is used since the value of t ₁ is 1. Since jmax is 1, the value of t ₂ is equal to the value of the element C _{2, 2.} Therefore, t ₂ = C ₂₂ = 0 − (14).

As the value of t ₃ , the case of (ii) above is used since the value of t ₂ is 0. Since jmin is 2, the value of t ₃ is equal to the value of element C ₃₂ . Therefore, t ₃ = C ₃₂ = 1- (15).

As the value of t ₄ , the case of (i) above is used since the value of t ₃ is 1. Since jmax is 2, the value of t ₄ is equal to the value of element C ₄₃ . Therefore, t ₄ = C ₄₃ = 0 − (16).

As described above, from the expressions (13), (14), (15), and (16), the projection vector T ₆ is T ₆ (t ₁ , t ₂ , t ₃ , t ₄ ) = (1010).

By the way, the elements used when obtaining the projection vector T ₆ are C ₁₁ , C ₂₂ , C ₃₂ , and C ₄₃ . Other elements C ₁₂ ,
C ₁₃ , C ₁₄ , C ₂₃ , and C ₃₄ are always 0, and need not be calculated. Furthermore, the elements of the projection matrix used when obtaining the projection vector T ₆ are C ₁₁ , C ₂₂ , C ₃₂ , and C ₄₃ so that the elements of the projection matrix required when obtaining the elements of the projection vector are From (i) and (ii), only those having an index of jmax or jmin are not involved. Therefore, it is not necessary to calculate other elements. That is, when obtaining the projection vector T, some elements of the projection matrix C are unnecessary. This is effective in simplifying the circuit.

In the above description, the dimension of the projection vector is 4, but the above method is expanded to have a general dimension of l as follows.

It has been mentioned above that the _first element t ₁ of the projection vector T (t ₁ , t ₂ , ..., T _l ) is equal to the projection matrix element C ₁₁ . Here, when the value of the element C ₁₁ is 1, the value of the elements C ₂₁ , C ₃₁ , ..., C ₁₁ in the first column is always 1. This is because the points P ₂ , P ₃ , ..., P ₁ always exist above the point P ₁ . If Cij is 1, element Ci + in the j-th column
The values of 1, j, Ci + 2, j, ... C _l , j are always 1 as in the above case.

Also, the slope n / m of the projected straight line is equal to 1 or 1
Since it is smaller, each time the row number of the projection matrix increases by 1, the number of elements having a value of 1 in the row increases by 1 at most. That is, when the element of the projection matrix that determines the i-th element ti of the projection vector is Cij, the following is shown.

When ti = Cij = 1, ti + 1 = Ci + 1, j + 1- (17) ti = Ci, j = 0, ti + 1 = Ci + 1, j- (18) From this, the projection vector T (t ₁ , t ₁ in FIG. t ₂ , ……,
t _l ). t ₁ is equal to the value of C ₁₁ . Therefore t ₁ is
It is 1. From the equation (17), t ₂ is equal to the value of C ₂₂ .
Therefore, t ₂ is 0. From Equation (18), t ₃ is equal to the value of C ₃₂ . Therefore t ₃ is 1. t ₄ is calculated from equation (17)
Equal to the value of C ₄₃ . Therefore, t ₄ is 1. The above process is repeated.

The element of the projection matrix that determines the value of t _l-2 is t _{l-1.
If it is l-2, k-2} and its value is 0, then from the equation (18), C
_If the element of the projection matrix that determines the value of _{l 1-2} is equal to the value of 1−1 _{, k−2} and is C _{1−2, k−2} , and the value is 1, then the first (1
From the equation (7) _, it is equal to the value of C _{l−1, k−1} .

Similarly, the value of t is equal to the value of C _{l, k−1} when t _l−1 = C _{l−1, k−1} = 0, and t _1-1 = C _{l−1, k− 1} = 1
, Then equal to the value of C _{l, k} .

Next, a projection vector T (t ₁ , t ₂ , ...
, T _m ), there are two methods of (a) expanding the above method into m dimensions and (b) sequentially obtaining l pieces each. In (a), since only 1 = m is set, the case of (b) will be described below. In order to simplify the description, the case of l = 44 (K-1) <m ≦ 4K will be described.

In FIG. 7, elements C ₁₁ , C ₂₁ , C _{22 of} the projection matrix,
C ₃₁ , C ₃₂ , C ₃₃ , C ₄₁ , C ₄₂ , and C ₄₄ are calculated, and the projection vector T (t ₁ , t ₂ , t ₃ , t ₄ ) is obtained. Next projection vector T
When calculating the elements of the projection vector C necessary for obtaining (t ₅ , t ₆ , t ₇ , t ₈ ), a residual R ₄ having the same meaning as the initial residual R ₀ is required. .

The value of the residual R ₄ is five values obtained by the following equation (19), It is either That is, the initial value of the residual required to obtain the next four is the final step of the preceding stage. It is C ₄₁ , C ₄₂ C ₄₃ , and C ₄₄ that determine the next residual. this is,
There is the following correspondence (formula (20)).

When C ₄₁ = 0, R ₄ = R ₀ ＋ 4n When C ₄₁ = 1 and C ₄₂ = 0, R ₄ ＝ R ₀ ＋ 4n−m When C ₄₂ = 1 and C ₄₃ = 0, R ₄ ＝ R ₀ ＋ 4n−2m When C ₄₃ = 1 and C ₄₄ = 0, R ₄ ＝ R ₀ ＋ 4n−3m When C ₄₄ = 1 R ₄ ＝ R ₀ ＋ 4n−4m As shown above, the residual is calculated and the next set of projection matrix is obtained. The element of C is calculated to obtain the 4-bit projection vector T.

When obtaining the m-bit projection vector T, as shown in FIG. 6, the m-bit is divided into blocks each having 4 bits, and the above process is repeated k times until a projection vector having m elements is determined. I will repeat.

The above method can be easily extended to the case of obtaining 1 bit at the same time.

Further, in the case of enlargement, in FIG. 7, the arrangement of the result pixels in the x-axis direction and the arrangement of the original pixels in the y-axis direction may be assumed, and the projection vector calculation method and the projection vector do not change. That is, in the case of enlargement, it can be considered that the result pixel m bits correspond to the original pixel n bits (m ≧ n).

Next, the initial residual R ₀ , the original pixel is m bits, and the result pixel is n.
FIG. 5 shows a flow for simultaneously calculating 1 element of a projection vector consisting of m bits in the case where the original pixel is reduced to the result pixel in bits.

41 and 42 set 0 and R ₀ as initial values in the original pixel counter and the residual register. 43 is the number of original pixels m, initial residual
A projection matrix generation process for obtaining a projection matrix of l rows and 1 columns from R ₀ and the result pixel number n according to the formula (5) and a residual candidate generation process for obtaining l + 1 types of residuals according to the formula (19) .

44 is the first (1
This is a projection vector determination process for determining the elements of l projection vectors based on the equations (7) and (18).

45 denotes the residuals required to calculate the elements of the next l projection vectors, the elements of the projection matrix also obtained by the projection matrix generation processing of 43 from the residual candidates obtained by the residual candidate generation processing of 43. Is a residual determination process that is performed according to the equation (20) using. 46 is a process for increasing the original pixel counter and the projection information pointer by 1. 47 is a process of determining whether or not the calculation of all the elements of the projection vector has been completed. As described above, the m-bit projection vector is obtained.

Next, an actual device using the above principle will be described. FIG. 1 is a module connection diagram showing an embodiment of an arithmetic unit for generating a four-dimensional projection vector. 001, 002, 003, 0
04 is a carry calculator (hereinafter referred to as a shifter) that outputs 2n, 4n, 2m, and 4m, respectively. 300 and 302 are 3n,
It is an adder that calculates 3m. 501, 502, 503, and 504 are signal lines for n, m, R, and the next residual obtained in the sixth processing, respectively. Reference numeral 701 is a register for the next residual obtained in the sixth processing.

In the above configuration, as described above, since the projection vector does not change between enlargement and reduction, in order to simplify the description, the following will be described in the case of reduction of magnification n / m (m ≧ n).
This operation will be described. First, when the residual R and the magnification n / m are given, the first to fourth
N, m, and R are input to the arithmetic units I to IV, respectively.

First, in the first arithmetic unit I, R + n is added by the adder 101, the output thereof is compared with m by the comparator 201, and the element C _{11 of the} projection matrix is output.

In the second operation unit II, 2 from shifter 001, 003
n and 2m are calculated, and the adder 201 calculates R + 2n. M, 2n + R, and 2m by the comparators 211 and 212
And the magnitude of 2n + R are compared, and the elements C ₂₁ and C _{22 of the} projection matrix are calculated.

In the third arithmetic unit III, 3n and 3m are output by the adders 300 and 302, R + 3n is calculated by the adder 301, and m and R + 3n, 2m and R + 3n and 3 are calculated by the comparators 311, 312 and 313.
The magnitude comparison of m and R + 3n is performed. As a result, the elements C ₃₁ , C ₃₂ , and C _{33 of the} projection matrix are calculated.

In the 4th arithmetic unit IV, from shifter 002,004, 4
n and 4m are calculated respectively, and the adder 401 calculates R + 4n. Subtractors 411, 412, 413, 414 use R + 4n-m, R
+ 4n-2m, R + 4n-3m, and R + 4n-4m are calculated.
These are the values that may be the residuals in the next phase, and when this calculation is performed, the elements of the projection matrix are
C ₄₁ , C ₄₂ , C ₄₃ and C ₄₄ are also calculated at the same time.

The contents of the projection vector selection unit 601 are shown in FIG. When the elements of the projection matrix determined in the first to fourth arithmetic processes are input, the matrix circuit outputs a 4-bit projection vector corresponding thereto. This circuit is simplified from the property of the estimation matrix C described above. That is, the property that the elements of the projection matrix required to obtain the elements of the projection vector are only those having an index of jmax or jmin, and the number of elements having a value of 1 in a certain row is It is a property that only increments by 1 for each increment of 1. This makes the projection vector 4
This means that the number of elements of the projection matrix required to obtain bits is at most four.

Next, the content of the residual selection unit 603 is shown in FIG. Elements C ₄₁ , C _{42 of} the projection matrix determined by the fourth arithmetic unit IV,
C _43, the content of _{C 44 R + 4n, R +} 4n-m, R + 4n-2
One of m, R + 4n-3m, and R + 4n-4m is output as the residual to be used next. This circuit is also simplified by utilizing the above property.

In FIG. 1, the path from the inputs n, m, and R to the output of the 4-bit projection vector passes through the carry arithmetic unit 001, the adders 300 and 301, and the comparator 311. Sixth
Since the longest one passes through 601 and 602, the projection vector can be obtained at high speed.

FIG. 4 shows a module diagram for obtaining an l-dimensional projection vector corresponding to FIG. In this way, it is possible to easily expand when 1 is obtained at the same time.

As described above, the present invention has a plurality of elements at the same time,
By obtaining the projection vector for controlling the scaling of the image, the time required conventionally can be greatly reduced.

[Brief description of drawings]

FIG. 1 is a module connection diagram of an arithmetic unit for a projection information parallel generation method according to an embodiment of the present invention, FIG.
FIG. 3 is a block diagram of a main part of FIG. 1, FIG. 4 is a diagram of a module diagram when the configuration of FIG. 1 is expanded to 1 bit, and FIG.
FIG. 6 is a flow chart for explaining the flow of the projection information parallel generation method in one embodiment of the present invention, FIG. 6 is a conceptual diagram showing the flow of residuals, and FIG. 7 is for generating a four-dimensional projection vector of the present invention. 8 is a conceptual diagram showing the concept for
Conceptual diagrams when expanded to bits, FIGS. 9 and 10 are conceptual diagrams showing how to obtain the elements of the projection vector using jmax and jmin, respectively, and FIG. 11 is for generating conventional projection vectors. FIG. 12 is a conceptual diagram showing a basic concept of FIG. 12, FIG. 12 is a conceptual diagram showing an example of a conventional reduction ratio of 3/7, and FIG. 13 is a flow chart showing a flow of a conventional method. 001, 002, 003 ... Carry calculator, 101, 201, 301, 401 ...
… Adder, 201, 211, 212, 311, 312, 313 …… Comparator,
411, 412, 413, 414 ... subtractor, 601 ... projection vector selection unit, 603 ... residual selection unit, 701 ... residual register.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Katsura Kawakami 3-10-10 Higashisanda, Tama-ku, Kawasaki City, Kanagawa Prefecture Matsushita Giken Co., Ltd. (56) Reference JP-A-60-189579 (JP, A)

Claims (1)

[Claims] 1. A L-bit (L is J, which represents a correspondence relationship between a first pixel column consisting of J pixels and a second pixel column consisting of K pixels through a mapping to a straight line. A projection vector consisting of a large number of K) is set to P from the n, m that determines the reduction ratio n / m <1 between the first pixel row and the second pixel row, and the given quantization residual. A projection matrix generation process for obtaining the elements of the projection matrix of row P columns (L> P) in parallel, and P + 1 candidates of quantized residuals used in the next projection matrix generation process subsequent to the projection matrix generation process are n , m, and a residual candidate generation process that is obtained in parallel from the quantized residuals used in the projection matrix generation process, and P from the projection matrix obtained by the projection matrix generation process.
Projection vector determination processing for determining the elements of the projection vectors in parallel, P + 1 residual candidates generated by the residual candidate generation processing, and P rows and P columns obtained by the projection matrix generation processing. A process for determining in parallel P bits (L> P) at a time, which consists of a residual determination process for determining a quantized residual used in the next projection matrix generation process from the elements in the Pth row of the projection matrix, is performed Q times. (Q ≧ 1) A projection information parallel generation method characterized by being obtained by performing.