JP2713400B2 - Projection information generation device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a projection information generating apparatus that calculates information for controlling image information scaling processing by a computer. 2. Description of the Related Art To enlarge or reduce an image, an original pixel {Av, w} is converted into a result pixel {Bx,
This is the process of converting to y｝. That is, this is a task of calculating the indices x, y of the result pixels that should correspond to the indices V, W of the original pixels, and associating the value of {Av, w} with {Bx, y}. In this enlargement / reduction processing, there is a method of obtaining a result pixel {Bx, y} by mapping the original pixel {Av, w} to a straight line. Hereinafter, this will be described. For simplicity of description, m bits of a one-dimensional original pixel are reduced to n bits of a result pixel (m ≧
n) will be described. FIG. 6 shows m original pixels, n
FIG. 8 is a diagram for describing an example of a principle of reducing the number of result pixels into pixels. (A ₁ , a ₂ ,..., A _m ) is the original pixel, (b ₁ , b ₂ ,
.., B _n ) are the result images. Hereinafter, for simplicity, the vectors (a ₁ , a ₂ ,..., A _m ) are denoted by A, and (b ₁ , b ₂ ,..., B _n) are _denoted by B. In FIG. The arrangement of A is assumed to be the arrangement of the result image B in the y-axis direction, and the horizontal grid line at the boundary between the result pixels b _j-1 and b _j is assumed to be h _j . the rate n / m is the slope, a value representing an error of the quantization (hereinafter referred to as residual) the initial value r _o of a line having the y intercept. grid line spacing is 1 for the sake of simplicity of explanation . resulting image B is the original image a as mapping through the straight line L, passes can be obtained. a perpendicular line passing through the center of the original pixels a _1, a. point P ₁ to the intersection of the straight line L and P ₁ When the results pixels horizon penetrates and b _1, i is assumed for j,
transferring the value of a ₁ to b _j. a ₁ is a method for obtaining the results pixels corresponding to neighboring pixels a _{i + 1} of a ₁ when it is known in advance and corresponding b _j can be simplified. That is, a _{i + 1} corresponds to either b _j or b _{j + 1} , which is determined by a horizontal grid line h _{j + 1} between point P _i and point P _{i + 1.} Can be determined by ascertaining whether there is. As described above, the correspondence between the original pixel and the result pixel can be obtained by the method of mapping onto the straight line L, and the information of the correspondence is expressed by the vector T (t ₁ , t ₂ ,..., T _m ). Can be represented by (Hereinafter projecting the vector T vector, referred to the elements and projection information.) Projection vector T is the projection information t ₁ is the element "1"
Alternatively, it is a binary vector taking any value of “0”. The value is determined by the following procedure. When the value of the projection information t ₁ becomes "1", the upper point P ₁ is a horizontal grid line h _1, or a when present on the horizontal grid lines h _1, the value of t ₁ is "0 "Is the other case. Ichi担, if a ₁ and b ₁ are found to correspond, a _{i + 1} corresponds to either of b _j or b _{j + 1.}
When the horizontal grid line h _{j + 1} exists between the points P ₁ and P _{i + 1} , or when the point P _{i + 1} exists on the horizontal grid line h _{j + 1} , a
_{i + 1} corresponds to b _{j + 1} . At this time, the value of t _{i + 1} is “1”. The value of t _{i + 1} becomes “0” in other cases. For example, a case where eight original pixels are reduced to three result pixels will be described with reference to FIG. The original image is A (a ₁ , a ₂
... a _3), and the result image _{B (b 1, b 2,} b 3). The inclination of the straight line L in the figure is 3/8. R ₀ which is the y-intercept of the straight line L is
This is an initial value of a value representing a quantization error. The intersection of the perpendicular and the straight line L passing through the center of the original pixels P _1, P ₂ ... from the left, respectively, and P _8. The value of t ₁ is the point P ₁ becomes "1", otherwise "0" when positioned above the horizontal grid lines h _1. In Figure 9, the point P ₁ is located on the lower horizontal grid lines h _1. Therefore, t ₁ = 0. Next, the value of t ₂ is determined. After the second element of T,
As described above, the previous intersection (in this case, P ₁ is a) between, on whether the horizontal grid lines are present, determined value. Between points P ₁ and point P _2, are present horizontal grid lines h _1. Therefore, t ₂ = 1. As described above, the projection vector T becomes T = (01001010) as far as t ₈ is obtained. The projection vector T calculated as described above indicates the correspondence between the original image A and the result image B based on the principle of the calculation. The image enlargement / reduction process calculates the projection vector T. There is nothing else. For example, it is possible to obtain a reduced pixel by transferring only the original pixel having the same suffix as the element having the value of “1” to the result pixel among the projection information that is the element of the projection vector T. In the case of enlargement (n> m), the original image is represented by B (b ₁ ,
b ₂ ,..., b _n ) and the resulting image as A (a ₁ , a ₂ ..., a _m ), and the projection vectors are the same. or,
When the original pixels are two-dimensional, the processing in the vertical direction may be performed after the processing in the horizontal direction, or the processing in the horizontal direction may be performed after the processing in the vertical direction. Calculation of the projection vector, the intersection P ₁ of the perpendicular and the line L drawn from the center of the original pixels a _1, a computation of quantizing to correspond to the lattice points, which the conventional DDA Algorithm 1
It has been determined by sequential processing using a plurality of arithmetic units. The actual calculation process will be described below. If the value of the y coordinate of point P ₁ is y ₁ , the residual at this point
r _i is r _i = y _{i −i} . However _i is the height of the horizontal line h _i directly below the point P _i. If the reduction ratio n / m is the length between the grating lines and m, the residual R _i = m · r _i, a value obtained by adding n to the comparison between m performed. (In the following description, the residual when the length between grid lines is set to 1 is a lowercase r, and when the original pixel is multiplied is an uppercase R.
Shall be used. ) When R _{i + n} is greater than or equal to m, the value of t _{i + 1} which is the (i + 1) -th projection information is set to 1
The value of the next residual _{Ri + 1} is defined as Ri _{+ n-} m. If smaller, the value of t _{i + 1} is set to 0, and the value of the next residual R _{i + 1} is set to R _{i + n} . (For details on this algorithm, see K. Kawakami and
S.Simazaki. “A Special Purpose LSI Processor U sin
g The DDA Aigorithm For Lmage Transformation ", (IE
EE, ComArc, llth, 1984 pp 48-54. Problems to be Solved by the Invention In the conventional method, since the generation of the projection information is performed sequentially by a program, a large number of calculations are required, and the speed of the enlargement / reduction processing is increased. Was difficult. In view of the above problems, the present invention simplifies arithmetic processing,
It is intended to generate projection information required at the time of enlargement / reduction of image information at high speed. Means for Solving the Problems In order to achieve the above object, the present invention provides a one-dimensional first image composed of m pixels and a one-dimensional second image composed of n pixels. L bits representing the corresponding relationship through Utsushikage to the straight line (L is, m, a large number of n) projection vector of, m, n, and, Q bits (Q from the initial residual R _O is m, (a small number of n) in the projection vector, the two consecutive run lengths P and P-1 representing the appearance cycle of "1" and "0" are calculated, and the run length is calculated by calculating the run length. The processing to be generated in units is first performed by “1” in the two types of run lengths P and P−1 and the projection vector from m, n and the initial residual R _O.
The first run l ₁ , which is the number of elements until the element of
P = [L / Q] +1, q = Q−R <L / Q>, l ₁ = P
- [(q + R _O) / Q], Y = R <(q + R O) / Q> ( symbols R <
L / Q> represents the remainder of division between integers where L is the dividend and Q is the divisor, and the symbol [X] represents the largest integer that does not exceed X). In the case where Y + q ≧ Q based on q and Y input from the first calculating means,
A value representing the appearance of the run length P-1 is used as an element of the run vector, and the value of Y is updated as Y = Y + q-n. Element and Y
A second computing means for calculating a continuous vector by repeating the updating of the value of Y as Y = Y + q Q-1 times;
Third calculation that synthetic projection vector and communicating vector calculated by the calculation means and the first of the two run-length P calculated by the calculating means, P-1, and the first from the run length l ₁ It is realized by means. Operation Since the present invention obtains the run length of projection information by the above configuration, it is possible to simultaneously calculate a plurality of bits of projection information which has been conventionally calculated for each bit, thereby reducing the calculation time conventionally required. It is. Embodiment As described above, the process of enlarging or reducing a two-dimensional image is an operation of converting an image {A _{v, w} } into an image {B _{x, y} }. This process results in a reduction of the one-dimensional area, regardless of the magnification and the size of the area. Hereinafter, for the sake of simplicity, a description will be given by taking a reduction of a one-dimensional area as an example. Here, some terms and symbols are defined. The number of original pixels is m, the number of resultant pixels is n, the initial value of the residual representing the quantization error is r _o , and the correspondence between the original pixels and the resulting pixels is defined as a pixel in which the pixels constituting the original pixels constitute the resultant pixels A vector T made up of m elements, in which the case where they appear and the case where they do not appear, is represented by 1 and 0, respectively.
<M / n> is a function representing the remainder of division between integers where m is the dividend and n is the divisor, and the symbol [X] is a function representing the largest integer not exceeding X. From the relationship m = n [m / n] + R <m / n>, P = [m / n] + 1− (1) q = n−R <m / n> − (2) , P, q satisfy m = nPq- (3). First, the basic concept of the present invention will be described. As described above, the projection vector T for projecting m one-dimensional original pixels onto n one-dimensional result pixels is composed of m elements. This is represented by a vector T (t ₁ , t ₂ ..., T _m ). Among the elements of T, those having a value of “1” are equal to the number of result pixels, and there are n elements. Among the elements of the projection vector T having the value of “1”, the K-th element is represented as t _{i (K)} . T _{i (1) which first} takes a value of “1” in the projection vector T
Is the minimum integer X that satisfies
Is nothing more than calculating. That L = (n / m) X + r o ≧ 1 - (4) from X ≧ (m / n) ( 1-r o) = m / n- (m / n) · r o = (nP-q) / n− (m / n) · _ro = P− (q + m · _ro ) · n− (5) Further, since X to be obtained is the minimum integer, m · _ro =
Let _Ro be i ₍₁₎ = [X] = P-[(q + _m.ro ) / n] = P-[(q + _Ro ) / n]-(6) Then, ti _(K-1) is In the case of “1”, i _(K−1) is (K ≧ 2) (n / m) · i _(K−1) + _ro = α + rK ₋₁ where α is an integer and rK ₋₁ is 0 ≦ r _K−1 <n / m｝ − (7) is satisfied. In order for t _{i (K) to} be “1”, X = i _(K) −i _(K−1), and the minimum integer satisfying L = (n / m) · X + r _K−1 ≧ 1− (8) It is nothing but calculating X. That is, X ≧ P − ((q + m · r _K−1 ) / n) according to Equations (4) and (5) − (9) From Equation (7), 0 ≦ m · r _K−1 = R _K−1 < n− (10) From Expression (2), 0 <q ≦ n− (11) Therefore, by adding Expression (10) and Expression (11) and dividing by n, 0 <(q + R _K−1 ) / n <2- (12) Since X is the minimum integer satisfying the expression (9), X = P: (0 <(q + _RK-1 ) / n <1)-(13) X = P-1 : (1 ≦ (q + R _K−1 ) / n <2) − (14) That is, i _(K) −i _(K−1) is P or P−1− (15) Therefore, t _{i (K On} the left side of (K ≧ 2), P−2 or P−1 “0” s are arranged as shown in FIG. As shown in the figure, a set of elements from 0 to ti _(K) immediately to the right of t _{i (K−1)} is referred to as a “run”, and the number of elements of the “run” is referred to as a run length l _K. FIG. 6 is an example of a projection vector when m and n are 37 and 9, respectively. In FIG. 6, the value of P is 5. In the example shown in the figure, it can be seen that the run length is 5 or 4 for the fourth and subsequent elements from the left. Which of the run lengths can be determined by calculating as follows. That is, from equations (13) and (14),
Using the residual _RK-1 at the time of calculating t _{i (K-1)} , if q + R _K-1 <n, then l _K = P q + R _K-1 ≧ n, then l _K = P-1 − (16 ) also, necessary for calculating the t _{i (K + 1),} the residual R _K, when the equation (3) _{(i) l K = P L} = (n / m) · P + R K-1 / m = (n / m) · {(m + q) / n} + R K-1 / m = 1 + (q + R K-1) / m ∴R K = q + R K-1 (ii) l K = P-1 of In the case, L = (n / m) · (P−1) + R _K−1 / m = (n · p−q + q−n) / m + R _K−1 / m = m / m + (q + R _K−1 −n) / _{m ∴R K = q + R K} -1 q + R K-1 by more -n <if n, l _K = P in _{R K = q + R K-} 1 - (17) if _{q + R K-1 ≧ n} , l K = At P-1, R _K = q + R _K−1 −n − (18) That is, the K-th (K ≧ 2) run length l _K is obtained by adding q to the residual R _K−1 .
Is added, and the result is compared with n to determine whether the number of elements is P or P-1. Further, the residual R ₁ for calculating the l ₂ are calculated as follows. That is, from equation (6), L = (n / m) (P-[(q + _Ro ) / n]) + _ro = (nPq + qn [q + _Ro ] / n]) / m + _Ro / m = [n · P−q) / m + [(q + _Ro− n · [(q + _Ro ) /
n]} / m = 1 + {q + R o -n · [(q + R _o) / _{n]} / m ∴R 1 = q +} R o -n · [(q + R _o) / n] = R [(q + R _o) / n − (19) Next, the following vector G (g ₂ ,..., G _n ) is defined,
Hereinafter, it is referred to as a continuous vector. That is, _gk is (i) _gk = 0 when RK _-1 + qn <0. (Ii) _gk = 1- (20) when RK _-1 + qn≥0. _For the value of _k and the run length l _K ,
According to (18), the following relationship exists. That is, (i) When g _k = 0, l _K = P, _RK = R _K-1 + q (ii) When g _k = 1, l _K = P-1, _RK = R _K-1 + q-n- (21) As described above, by calculating one g _k , which is an element of a continuous vector, P elements or P−1
Calculation can be performed for each individual. The run length l1 of the _first run (hereinafter, referred to as the initial run) is calculated based on Expression (6). FIG. 5 is a flowchart of a specific method for obtaining a projection vector using this principle. In step 101, processing for inputting parameters m, n, and _Ro is performed. In 104, run lengths P and q
Is calculated based on the equations (1) and (2). The initial run length l ₁ in 105 is calculated based on the equation (6). Further, the residual Y of the element having the first value of “1” of the projection vector is calculated based on the equation (19). At 106, a process of outputting an initial run composed of l _one element calculated at 105 is performed. In 107, a process of adding the variation amount q of the residual to the residual Y is performed, and the next residual y before being changed to a value of 0 or more and less than n is obtained. At 108, a process of comparing the magnitudes of y and n is performed. y is equal to n or n
If it is larger, a process of outputting a run having a run length of P-1 at 110 is performed. In other cases, a process of outputting a run having a run length of P in step 109 is performed. In step 111, when P-1 runs are output, the residual necessary for the calculation of the next run length is defined as y-n. At 112, it is determined whether m pieces of projection information have been obtained. The projection vector T is calculated as described above. Next, an embodiment of the projection vector generation device based on the concept of the present invention will be described. FIG. 1 is a block diagram of a projection vector generating apparatus according to an embodiment of the present invention. In the figure, the first arithmetic unit 1 has run lengths P, P
-1, it is calculated parameters required in the second arithmetic unit which will be described later with l ₁ n, q, and Y. That is, the processing of 101 to 105 in FIG. 5 is performed. The second arithmetic unit 2 performs a process of calculating a continuous vector. That is, the processing of steps 107, 108 and 111 in FIG. 5 is performed. Third arithmetic unit 3, run-length P calculated by the first arithmetic unit 1, and P-1, l _1, from the communication vector calculated by the second arithmetic unit 2, and outputs the continuous projection information. That is, the processing of steps 106, 109, 110, and 112 in FIG. 5 is performed. The projection information register 4 stores 32-bit projection information output by the third arithmetic unit 3. Hereinafter, a more detailed configuration of FIG. 1 will be described.
FIG. 2 is a detailed block connection diagram of the first arithmetic unit 1. Hereinafter, the operation will be described. Reference numeral 207 denotes a divider having m, n, and _Ro as inputs, performs the following processing, and outputs P, q, l ₁ , and Y. P = [m / n] + _1- (22) q = n-R [m / n]-(23) ₁₁ = P-[(q + _Ro ) / n]-(24) Y = R [(q + _Ro ) / N] − (25) The subtractor 204 outputs P−1− (26). The first arithmetic unit performs the above processing. Calculated
Y, q is an input of the second operational unit, P, is P-1, l ₁ is the input of the third arithmetic unit. FIG. 3 is a block diagram of the second operation unit 2, the third operation unit 3, and the projection information storage register 4. 30
1 to 306 are Y, q, n, P, P−1, l ₁ calculated by the first arithmetic unit.
Are stored in the register. 310 and 312 are adders, and 311 and 313 are subtractors. Reference numeral 308 denotes a code register that stores code information of the operation result of the subtractor 311. A carry register 309 stores carry information of the operation result of the adder 309. Reference numerals 321 to 324 denote selectors with two inputs and one output. The selectors 321 to 322 are the sign register 30
The selector 323-324 is controlled by the value of the carry register 309. 333 is a buffer for controlling timing. Reference numerals 341 to 344 denote AND gates, and reference numerals 345 to 347 denote NOT gates. 351 is a decoder. 300 is a 32-bit register. A register 352 stores the operation result of the adder 312. Φ1, φ2, φ in the figure
3 is a clock, and FIG. 4 is a diagram for explaining the phase relationship between these clocks. As in the figure, these three clocks, constituting the period of T ₁ through T _6. Hereinafter, the operation will be described. When the processing by the first operation unit 1 is completed, the calculated Y, q, n, P, P-1, l 1 is stored in the register 301 to 306. All other registers store the initial value “0”. (1) Processing 1 (i) register with a period of T ₁ 301,306,307 is updated. (Not updated in the period of the first T _1, the respective registers, Y, l _1, 0 is stored) (ii) a period of T _2, the process of the adder 310 of the second arithmetic unit , The processing in the subtractor 311 and the processing in the adder 312 of the third operation unit. The adder 310 outputs Y + q− (27). The result of this operation is supplied to the selector 321 and the subtractor 311.
Enter. The subtractor 311 outputs (Y + q) -n- (28). The value of the sign register 308 is as follows: S = 1; (Y + q) −n ≧ 0− (29) S = 0; (Y + q) −n ≧ 0 That is, in the period of T ₂ , the process of calculating one bit of the element of the continuous vector and storing it in the code register 308 is performed. The adder 312 adds the value B of the register 306 and the value L of the register 307, that is, outputs B + L− (30). When the calculation result is 33 or more, "1" is output as carry information. In other cases, the carry information to be output is “0”. The subtractor 313 outputs (B + L) −32− (31). (Iii) T register 309 in the period of _3, when the value C of the updated (final) register 309 is 1, processing 2 is executed, when C is 0, the processing 3 is executed. Processes 2 and 3 are
It is shown below. (2) through the AND gate 344 for a period of treatment 2 (i) T _4, register 30
There is a clock at 0,004. With this clock, the register 004 stores the value of the register 300. Then register
All 300 are set to “0”. In period (ii) T _5, the register 352 is updated, NOT
The value D is not input to the decoder 35 by the gate 346 and the AND gate 343. Also, NOT gate 345, AND gate 3
According to 42, the clock does not enter the register 308 and the register 308 is not updated. The selector 323 in the period (iii) T ₆ is the output of the subtracter 313 (B + L) -32 - outputs (32). The selector 324 outputs “0”. Register 306 and 307 for a period of (iv) T ₁ is updated. NOT
Due to the gate 347 and the AND gate 341, no clock enters the register 301 and the register 301 is not updated. (V) an adder 312 for a period of T _2, the subtractor 313, the process 1-
Perform the processing described in (ii). Register 309 in the period (vi) T ₃ is updated. (End) When the value C of the register 309 is “1”, the above (i) to
The processing of (vi) is performed until the value of C becomes “0”. (3) Since the processing 3 (i) register with a period of T ₄ 300,004 clocks not enter, there is no exchange of data between the registers 300,004. Register 382 in the period (ii) T ₅ is updated, the value D is Dekodda 351, it enters the selector 324. The clock is input to the register 308, and the register 308 is updated. Selector 321,3
32, when the value S of the register 308 is “0”, Y +
q and P are output, and when S is "1", Y + q-n,
Outputs P-1. (Iii) for a period of T _6, the decoder 351 decodes the value of the register 352 and outputs "1". The register 300 stores the output of “1” of the decoder 351. For example, when D = 6, the register 300 stores “1” in the seventh bit. The selector 323 outputs the input from the selector 322,
The selector 324 outputs an input from the register 352.
(End) When the process 3 ends, the process 1 is executed. By performing the above processing until all the projection information is calculated, the projection information can be calculated simultaneously for each of a plurality of bits. Effect of the Invention As described above, the present invention provides an L-bit (L bit) representing a correspondence between a first image composed of m pixels and a second image composed of n pixels via a straight line projection. L is, m, the projected vector of a large number) of n, m, n, and, Q bits from the initial residual R _o (Q is m, "in the projection vectors in of small number) of the m 1 By calculating a continuous vector representing the appearance order of two types of continuous lengths representing the appearance cycle of "," 0 ", it is possible to simultaneously calculate a plurality of bits of projection information which was conventionally calculated for each bit. Calculation time can be reduced.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of a projection information generating apparatus according to an embodiment of the present invention, FIG. 2 is a module diagram of a first arithmetic unit which is a main part of FIG. 1, FIG. FIG. 5 is a module connection diagram of the second and third arithmetic units and the projection information storage register, which are main parts of FIG. 1, FIG. 4 is a waveform diagram for explaining the phase relationship of clocks of main parts in the apparatus, and FIG. Is an operation flowchart showing the basic principle of the present invention, and FIG.
FIG. 6 is a conceptual diagram for explaining the A calculation, FIG. 6 is a diagram showing raw projection information, FIG. 6 is a conceptual diagram for explaining conventional scaling, FIG.
FIG. 9 to FIG. 9 are conceptual diagrams for calculating projection information at the time of conventional enlargement / reduction. DESCRIPTION OF SYMBOLS 1 ... 1st arithmetic unit, 2 ... 2nd arithmetic unit, 3 ... 3rd arithmetic unit, 4 ... Register, 204,311,313 ... Subtractor, 310,312 ... Adder, 207 ... Divider, 321-324 ... Selector, 004,300-309,352,353 ... Register, 351… Decoder, 341-344… AND gate, 345-347… NOT gate, 333
…buffer.

   ────────────────────────────────────────────────── ─── Continuation of front page    (72) Inventor Shigeo Shimazaki               Matsu, 3-10-1 Higashimita, Tama-ku, Kawasaki-shi               Shimogiken Co., Ltd. (72) Inventor Chika Onodera               Matsu, 3-10-1 Higashimita, Tama-ku, Kawasaki-shi               Shimogiken Co., Ltd.

Claims (1)

(57) [Claims] L bits (L is m, m, L) representing a correspondence between a one-dimensional first image composed of m pixels and a one-dimensional second image composed of n pixels via a straight line mapping. the projection vectors of large number) of n, m, n, and, Q bits (Q is the initial residual R _o m, during the projection vector of small number) of n "1", "0" two run-length P representing the occurrence period of, by calculating a continuous vector representing the order of appearance of P-1, two types of processing for generating in the continuous length unit, m, n, and the initial residual R _o , And the first run l ₁ , which is the number of elements until the first “1” element appears in the projection vector, and the residual Y, necessary to calculate the run vector, And the variation q of the residual is P = [L / Q]
+1, q = Q-R <L / Q>, l ₁ = P-[(q + _Ro ) / Q], Y =
R <(q + R _o ) / Q> (symbol R <L / Q> is L for dividend, Q
Represents the remainder of the division between integers with the symbol [X]
Represents the largest integer not exceeding X)
And Y + based on q and Y inputted from the first computing means.
When q ≧ Q, the value representing the appearance of the run length P−1 is used as an element of the run vector, and the value of Y is updated as Y = Y + q−n. When Y + q <Q, the appearance of the run length P is determined. The value to be represented is taken as an element of the continuous vector, and updating the value of Y to Y = Y + q is Q-
A second calculating means for calculating a run vector by repeating once, a run vector calculated by the second calculating means, and two types of run lengths P, P- calculated by the first calculating means.
1, and projection information generating apparatus characterized by being configured from a third arithmetic means for combining the projection vector 1st from run length l _1.