JP2849098B2 - Multivalue image reduction apparatus and method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Overview] The density of each converted pixel is calculated from the density of a neighboring converted pixel in accordance with a divided area in which the converted pixel is projected on a plane on which the converted pixel is arranged by a projection method. A multi-valued image reducing apparatus and method for determining a converted pixel projected on a plane on which a pixel to be converted is arranged, with the object of enabling a multi-valued image to be reduced without losing information of a high-frequency component In the multi-valued image reduction device that determines the density of each converted pixel from the density of the pixel to be converted in the vicinity thereof in accordance with the decomposition area based on the projection method, a predetermined converted pixel group located in the vicinity of the converted pixel is Converted pixel input means for inputting density data, thinned pixel detecting means for detecting original pixels to be thinned out at the time of reduction by the coordinates of the converted pixels and conversion magnification, the converted pixel input means and the thinned pixel detecting means Receiving the output of the high-frequency pattern detection means for detecting high-frequency components from the density data of the group of pixels to be converted, receiving the coordinates of the converted pixels and the conversion magnification, based on the array of the pixels to be converted, and within the area based on the array of the pixels to be converted To which of the divided regions of the divided regions set according to the conversion magnification as to contribute to the density of the converted pixel by each area ratio in which each density of the group of pixels to be converted has a size of a power of 2 corresponding thereto. A divided region determining unit that determines whether the converted pixel is located, and each of the converted pixels displayed in a binary system based on the divided region determined by the divided region determining unit and the detection data by the high-frequency pattern detecting unit. A plurality of digit shifting means for shifting the digit of the density data; and a converted pixel density calculating means for calculating the density of the converted pixel from the density data of each converted pixel shifted by the digit shifting means. Ri to configure.

[Industrial Application Field] The present invention uses the projection method to calculate the density of each converted pixel from the density of a neighboring converted pixel in accordance with a divided area where the converted pixel is projected on a plane on which the converted pixel is arranged. The present invention relates to a multivalued image reduction apparatus and method for determining.

In current image processing systems, multi-valued images such as halftones and full-color images as well as black-and-white binary images are increasingly being targeted, and enlargement / reduction techniques when editing multi-valued images are also indispensable. Is coming. In image input / output devices, halftone and full color scanners and printers (thermal transfer or electrophotography) have also become widespread, and a pixel density converter between devices having different resolutions plays an important role.

[Prior Art] A projection method used as a pixel density conversion device (enlargement / reduction device) for a multi-valued image is a method of determining the density of a conversion pixel from four nearest pixels and adjacent pixels to be converted. Electronic Journal, Vol. 11, No. 2, pp. 72-83,198
2). FIG. 20 is a diagram showing the principle of the projection method. In the projection method, which is one of the geometric mode conversion, four pixels A, B, C,
The converted pixel including D is divided into four equal parts as shown in the figure. Four neighboring pixels the density, respectively I _A of, I _B, I _C, and I _D, the converted pixel R
(Dots in the figure) on the pixel surface to be converted,
Area ratios projected to the quadrant are defined as W _A , W _B , W _C , and W _D. Then, determine the concentration I _R of the converted pixel _{R I R = ΣIi × Wi (} 1) i = A, B, C, the D. Here, generally, ΣWi = 1. In addition,
P and q in FIG. 20 are conversion magnifications in the X and Y axis directions.

FIG. 19 is a flowchart showing processing means of the projection method. First, the data of the pixel to be converted is inputted, and the coordinates (X, Y) of the converted pixel R on the pixel to be converted are calculated (step 1). Next, the respective area ratios W _A , W _B , W _C , and W _D projected from the coordinates (X, Y) and the conversion magnifications p, q are calculated (step 2). Next, the densities I _A , I _B , I _C , of the four pixels A, B, C, D to be converted
The density of the converted pixel R is calculated from _ID and each of the area ratios W _A , W _B , W _C , and W _D obtained in step 2 according to the equation (1) (step 3), and the density I _R is output (step 4). ).

FIG. 21 is a block diagram showing an area ratio and product-sum operation circuit by the projection method. This schematic is shown in Figure 2, Step 2,
Equivalent to 3. In the figure, 1 is the current conversion element R (second
(See Figure 0), the position coordinates (X, Y) and the conversion magnification p, q in each of the X and Y axes (see FIG. 20).
This is an area ratio calculation circuit that calculates each area ratio W _A , W _B , W _C , and W _D corresponding to the position of the converted pixel R projected from p and q. 2
5 are density data I _A , I _B , I _C , and I _D of the converted pixels located at the converted pixels A, B, C, and D shown in FIG. 20, and the area ratio is input to the other input. An acid multiplier that receives the area ratio data W _A , W _B , W _C , and W _D corresponding to each position output from the operation circuit 1 and performs the Ii × Wi operation represented by the equation (1).

6 is an adder for adding the outputs of the acid filters 2 and 3, 7 is an adder for adding the outputs of the acid filters 4 and 5, and 8 is an adder for adding the outputs of the adders 6 and 7. It is. The adder 8
Output the density data I _R of the converted pixel R shown in the equation (1). The area ratio calculation circuit 1 calculates the position coordinates of the converted pixel.
From the values of X and Y, their ± signs and conversion magnifications p and q, (0.
5 + pX) × (0.5 + qY) (2) is calculated to obtain each area ratio W _A , W _B , W _C , and W _D. Here, the signs of X and Y are determined based on whether X ≧ 0 to Y ≧ 0.
Each area ratio W _A , W _B , W _C , and W _D enters multipliers 2 to 5 and is multiplied by the density data I _A , I _B , I _C , and I _D, and each multiplication result is added to an adder 6 to
8, and finally output from the adder 8 as the density I _R of the converted pixel R shown in the equation (1). Therefore, in the circuit shown in the figure, multiplication is performed 16 times (3 × 4) +4 and addition / subtraction is performed 11 times (2 × 4) +3.

[Problems to be Solved by the Invention] In the conventional projection method as described above, since the density is averaged by the area ratio projected from the position of the conversion pixel, the image is blurred when the original image has high-frequency components. As a result, there is a problem that information of high frequency components is lost. 22nd
This will be described with reference to the drawings. The figure shows a case of 2/3 reduction processing by the conventional projection method. In the figure,
Pixels indicated by ○ are gray pixels close to white, pixels indicated by ● are gray pixels close to black, pixels indicated by × are converted pixels,
Consider a case where a high-frequency component (that is, a gray-scale image close to black) exists vertically in the center.

Now, consider the case of projecting the converted pixel R _1. The conversion pixels R _1, since the area ratio corresponding to the original pixel A (white near gray image of ○) is large, expressed in the form of an original pixel C, and information of D (black close gray images ●) was Usuma' . In the figure, the area ratio W _{A of A} is much larger than the area ratios W _C , W _{D of} D and C where ● is. Therefore, as a result of expressing the density in terms of the area ratio, the ● information fades. Since same is true for the converted pixel R ₂ to R _4, the concentration of the converted pixel R ₁ to R ₄ is it becomes uniform, the original pixel C, so that the information D would completely disappear.

The present invention has been made in view of such a problem, and an object of the present invention is to provide a multivalued image reducing apparatus and method capable of reducing a multivalued image without losing information of a high-frequency component.

[Means for Solving the Problems] FIG. 1 is a block diagram showing the principle of the device of the present invention (first invention), and FIG. 2 is a flowchart showing the principle of the method of the present invention (second invention).

First, FIG. 1 will be described. In the figure, reference numeral 11 denotes converted pixel input means for inputting density data of a predetermined group of converted pixels located in the vicinity of the converted pixel, and 12 detects original pixels which are thinned out at the time of reduction by the coordinates of the converted pixels and the conversion magnification. The thinned pixel detecting means 13 is a high frequency pattern detecting means which receives the output of the thinned pixel input means 11 and the thinned pixel detecting means 12 and detects a high frequency component from the density data of the group of pixels to be converted.

Here, the thinned pixel detecting means 12 detects an original pixel (thinned pixel) corresponding to the converted pixel determined by the conversion magnification when the converted pixel is located at a position where the information of the original pixel cannot be reflected. is there.

Numeral 14 receives the output of the high-frequency pattern detection means 13, the coordinates of the converted pixels, and the conversion magnification. Based on the array of the converted pixels, each density of the group of converted pixels corresponds to a power of 2 corresponding to the density in the area. A divided region determining unit that determines which divided region of the divided region set according to the conversion magnification has the converted pixel positioned therein as one that contributes to the density of the converted pixel by each area ratio having a size of Reference numeral 15 denotes a plurality of digits for performing digit shift of each density data of the pixel to be converted displayed in a binary system based on the divided area determined by the divided area determination means 14 and the detection data by the high frequency pattern detection means 13. The shifting means 16 are conversion pixel density calculating means for receiving the output of the digit shifting means 15 and calculating the density of the converted pixel from the density data of each converted pixel shifted by the digit shifting means 15.

Next, the second invention shown in FIG. 2 will be described. The present invention performs a process of sequentially inputting density data of a predetermined group of pixels to be converted located in the vicinity of a conversion pixel (step 1), and detects original pixels to be thinned out at the time of reduction by the coordinates of the conversion pixel and the conversion magnification. A step (step 2), a step of detecting high-frequency components from the density data of the group of pixels to be converted (step 3), and each density of the group of pixels to be converted corresponds to an area based on the array of the pixels to be converted. It is determined as to which divided area of the divided area set according to the conversion magnification the converted pixel is located, assuming that each area ratio having a size of a power of 2 contributes to the density of the converted pixel. (Step 4), and the density of each of the converted pixels displayed in a binary system based on the divided area determined in the step (Step 4) and the high-frequency component detection data in the step (Step 3). A step of shifting the digit of the data (step 5); and a step of calculating the density of the converted pixel from the density data of each converted pixel shifted in the step (step 5) (step 6). It is characterized by:

[Operation] The present invention detects a high-frequency component from multi-value information of an original image, and projects the information to a conversion pixel when the high-frequency component appears. It is intended to be preserved.

In step 1, based on a position where the converted pixel is projected on a plane formed of the pixel to be converted with respect to a predetermined conversion, the converted pixel input unit 11 is expressed in a binary system located near the converted pixel. The density data of the converted pixels are sequentially input to the pixel density conversion device. Here, A is the be converted pixels located in the vicinity surrounding the pixel R ₁ as shown in FIG. 8 for example, B, C, refer to 4 pixels and 4 pixels surrounding the D.

Next, in step 2, the thinned-out pixel detecting means 12 detects original pixels to be thinned out at the time of reduction using the coordinates (X, Y) of the converted pixel and the conversion magnifications p and q. In step 3, the high frequency pattern detecting means 13 detects a high frequency component from the density data of the eight pixels input by the pixel conversion pixel input means 11.

Next, in step 4, upon receiving the output of the high-frequency pattern detecting means 13, the divided area determining means 14 determines the divided area where the converted pixel is located. Here, the divided area is based on the conversion magnification and the array of the converted pixels in the case of the reduction of the pixel density conversion, and based on only the array of the converted pixels regardless of the conversion magnification in the case of the enlargement, and in the area, A region set to contribute to the converted pixel density by an area ratio having a size of a power of 2 with respect to the density of the converted pixel. Further, “contributing” is not necessarily limited to the one that contributes by an accurate area ratio determined by the position of the converted pixel, and may be treated as contributing if a certain tendency of contribution is considered.
The influence of the density of a non-converted pixel on the density of a certain converted pixel is expressed by the above-described equation (1). "Area ratio"
For example, the non-conversion pixels located near the conversion pixel are A, B,
When C and D are used, W _A , which is displayed as a coefficient of the density I _A , I _B , I _C , and I _D of each converted pixel with respect to the density I _R of the conversion pixel R,
W _B , W _C , and W _D.

Next, based on the divided area determined by the divided area determining means 14 and the output of the high frequency pattern detecting means 13 in step 5, an instruction to shift a digit is given to the plurality of digit moving means 15. Here, the number m of the digit moving means 15 indicates the number of pixels to be converted located in the vicinity of the conversion pixel, and m = 4 in a normal projection method. That is,
Each digit moving means 15 is provided for each pixel to be converted located in the vicinity of the conversion pixel, and independently shifts the digit for each pixel to be converted based on the high frequency condition of the divided region where the conversion pixel is located. Is done. As described above, since the influence of the density of each pixel to be converted on the density of the conversion element is made to correspond to each divided area at an area ratio carved every power of two, multiplication or the like is performed. This is because the effect of each pixel to be converted can be obtained only by moving the digit without performing the calculation.

Next, in step 6, the conversion pixel density calculating means 16 can calculate the density of the conversion pixel to be processed based on the influence of the density obtained for each of the converted pixels. For example, when the calculation of the density of the converted pixel is performed based on the above equation (1), only the addition is performed, and the calculation is greatly simplified.

[Example] Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 3 is a configuration block diagram showing one embodiment of the present invention. The same reference numerals as in FIG. 1 are used. In the figure, 12
Is a thinned pixel detection circuit, 13 is a high frequency pattern detection circuit,
Reference numeral 14 denotes a divided area determination circuit. 15a to 15d are digit moving means 15
Is a 4 → 1 multiplexer. The outputs of the multiplexers 15a and 15b enter the adder 16a,
The outputs of c and 15d enter an adder 16b, and these adders 16a and 16b
The output of the enters the adder 16c, the output of the adder 16c is density data I _R of the converted pixel R. The adders 16a to 16c correspond to the conversion pixel density calculating means 16 in FIG.

This embodiment is similar to the one shown in FIG. 8 which is located near a predetermined conversion pixel read by a conversion pixel input unit such as a transmission unit for transmitting data from a camera or a computer which reads data directly from an original image. Input each density data I _A , I _B , I _C , I _D, and I _{E to} I _L of four pixels, A, B, C, D and a pixel to be converted expressed in a binary system of four pixels around the four pixels. Multiplexers 15a to 15d corresponding to the digit shifting means 15 for shifting the digit (bit shift), and detecting the original pixels to be thinned out at the time of reduction from the coordinates of the conversion pixel currently being targeted and the conversion magnification. A thinned pixel detection circuit 12, a high frequency pattern detection circuit 13 for detecting high frequency components from density data of eight pixels input by the converted pixel input means 11, and a divided region determination circuit 14 for determining a divided region where a converted pixel is located And corresponds to the conversion pixel density calculation means 16. It is constituted by an adder 16 a to 16 c.

The thinned pixel detection circuit 12 detects thinned pixels based on the thinned data shown in FIG. 4 and the conditions shown in FIG.
FIG. 4 is an explanatory diagram of thinning data necessary for detecting a thinned pixel, FIG. 5 is a diagram showing four segmented areas, and FIG. 6 is a diagram showing a reduction ratio of 1> p, q ≧ 1/2 according to the embodiment. FIG. 10 is a diagram showing reference pixels in the case. As the thinning data, the conditions X ₀ , X ₁ , X ₂ ≧ 0? Y ₀ , Y ₁ , Y ₂ ≧ 0? And the skips between the four divided areas (see FIG. 5) of the converted pixel including the upper and lower and the front and rear Number of original pixels (xshift0,1 yshift
0,1), and the thinned pixels in the original image are detected based on the result (see FIG. 7).

Here, the thinned-out data in FIG. 4 is a condition for detecting a thinned-out pixel for which information is inevitably missing when the content of the thinned-out pixel is to be shown in a converted (reduced) pixel. When the image is reduced, the number of pixels to be converted is smaller than the number of pixels to be converted, so that there is a pixel to be converted farther than any of the converted pixels, and that information may be lost.

However, when the distance between the pixel to be converted and the converted pixel is strictly compared, the processing becomes complicated. Therefore, as shown in FIG. 4, the coordinates are provided based on the pixel to be converted, and the position (area) of the converted pixel and the positional relationship between the preceding and succeeding converted pixels are simplified. Is the purpose of the thinning data.

Specifically, the coordinates of the target conversion pixel R11 are set to (X1, Y
1). Four converted pixels A, B,
The coordinates are simplified by assuming that the center of the square formed by C and D is 0, 1 if it is positive in the horizontal (X) direction and vertical (Y) direction, and 0 if it is negative. Then, the former conversion pixel and the current conversion pixel, the current conversion pixel and the next conversion pixel to indicate the positional relationship, the former
xshift0 and yshift0 are used, and xshift1 and yshift1 are used for the latter. In this case, it indicates how many pixels to be converted exist between the converted pixels.

(Example) In the horizontal (X) direction: X coordinate X0 of previous conversion pixel R10 X coordinate X1 of current conversion pixel R11 X coordinate X2 of next conversion pixel R12 Relative positional relationship xshi between previous conversion pixel R10 and current conversion pixel R11
ft0 Relative positional relationship xs between current conversion pixel R11 and next conversion pixel R12
hift1 In the vertical (Y) direction X coordinate Y0 of the previous conversion pixel R01 Y coordinate Y1 of the current conversion pixel R11 Y coordinate Y2 of the next conversion pixel R21 Relative positional relationship yshi between the previous conversion pixel R01 and the current conversion pixel R11
ft0 Relative positional relationship ys between the current conversion pixel R11 and the next conversion pixel R21
hift1 Attempts to detect a pixel to be converted for which information is missing using the conditions above.

Next, a description will be given of the purpose of the four-part area shown in FIG. As described above, the above-described ten parameters are required to detect a pixel to be converted in which information is missing, and it is desired to reduce these parameters to further simplify the detection process. Therefore, the coordinates (X1, Y
The required parameters are reduced according to the position of 1), that is, the four divided areas in FIG.

For example, if a converted pixel exists in H1, X0, Y0, xshift
0, yshift0 is unnecessary and X1, X2, Y1, Y2, xshift1, yshift1
The pixel to be converted in which information is missing is detected by the six parameters. As an additional effect, the number of converted pixels to be referred to can be reduced. For example, as shown in FIG. 6, normally, it is necessary to refer to four lines of converted pixels. By limiting to the divided area, it is sufficient to refer to the converted pixels for three lines.

The reference pixel shown in FIG. 6 is a pixel to be converted for detecting the thinned data shown in FIG.

Next, (a) and (b) of FIG. 7 will be described. Here, that the condition is satisfied means that there is a pixel to be converted that is far away from the converted pixel and that information may be lost, and that the condition is not satisfied means that the converted pixel is close to the converted pixel. That is, the information of the pixel to be converted can be reflected without performing any special processing.

Here, the thin line patterns No. 1 to No. 4 shown as the target thin line patterns are the same as the thin line patterns shown in FIG. FIG. 7 shows a condition for obtaining a thinned pixel. If a thin line pattern as shown in FIG. 8 exists, a special process is performed by referring to the thinned pixel corresponding to this condition. .

Next, the connection with the high-frequency pattern detection processing after detecting the thinned pixel will be described. As shown in FIG. 9, two thresholds are provided for pixels having multi-value (multi-gradation) information, and the binarized pattern is the eighth pattern with either threshold.
When the thin line pattern shown in the figure is reached, it is determined that there is high-frequency information that may be thinned out, and the following high-frequency information storage processing is performed. The high-frequency pattern shown in FIG. 9 corresponds to the thin line pattern shown in FIG.

Here, “H1-−” described in the target fine line pattern is used.
2 "," H1-4 ", etc. This means that it overlaps with the thin line pattern of the immediately preceding converted pixel, and indicates that in this case, it is not the target of the thin line pattern. When the conversion pixel is in the condition of “H2-1”, the conversion pixel immediately before (in the y-axis direction) has already been the target of the fine line pattern under the condition of H1-2, so the current conversion pixel is the target of the fine line pattern. Otherwise, the thin line (1 pixel) becomes rather thick (2 pixels) because of the overlap.

The high-frequency pattern detection circuit 13 detects a high-frequency pattern from the input density data of eight pixels (FIG. 8). Two specific types of processing methods are used here.

(Process 1) As shown in FIG. 9, each density data of eight pixels is binarized by _two thresholds T ₁ and T ₂ , and each binarized pattern matches the previously prepared high-frequency pattern in FIG. Then, it is determined that the original pixel group has a high frequency component. No.
FIG. 10 is a flowchart showing the processing procedure. First, a target pixel to be converted (eight pixels) is input (step 1,
8 See FIG.) Is binarized by the threshold T ₁ (step 2). Binarized pattern Check has become a high-frequency pattern (Step 3), the If so then the threshold T ₂
(Step 4). Next, it is checked whether or not this binarized pattern is a high-frequency pattern (step 5).
If so, high frequency information storage processing is performed (step 6). If it is not a high-frequency pattern in steps 3 and 5, normal processing is performed (step 7).

(Process 2) As shown in FIG. 11, a match with the high-frequency pattern shown in FIG. 8 is first taken from the upper bit plane of each density data of eight pixels, and if they match, the pixel of the high-frequency component is removed. It is determined whether the four pixels are high-frequency data depending on whether the black and white values are the same until reaching the lower bit plane. FIG. 12 is a flowchart showing the determination process. First, a target pixel to be converted (eight pixels) is input (step 1), and N = 8 and T = 0 are initialized (step 2). Next, it is checked whether N> α (step 3), and if so, the N-th bit plane is selected (step 4). Here, if N> α, normal processing is performed (step 5).

Next, it is checked whether the bit plane selected in step 4 is a white pattern (step 6), and if so, N = N-1 is calculated and the process returns to step 3. If not, it is checked whether it is a high-frequency pattern (step 8), and if so, it is checked whether the colors of the outer four pixels are exactly the same (step 9). If it is not a high-frequency pattern in step 8, normal processing is performed. If the colors of the outer four pixels are exactly the same, N = N−1, T = T + 1 is calculated (step 10), the Nth bit plane is selected (step 11), and the process returns to step 9. .

If the colors of the outer four pixels are not exactly the same in step 9, it is checked whether T> β (step 1).
2) If so, the high-frequency information is stored (step 13).

Returning again to the description of FIG. In the divided area determination circuit 14, from the coordinate X, Y of the converted pixel and the conversion magnification p, q, the divided area shown in FIG. 13 and the division condition shown in FIG. Is determined. Fig. 15 shows 13
It is a flowchart which shows the area | region determination processing procedure of the figure. No.
As shown in FIG. 13, the area is divided into 13 areas G1 to G13. Then, depending on which of the division conditions shown in FIG. 14 is satisfied, an area to be obtained is obtained as shown in FIG.

Next, in the conversion pixel density calculation circuit 16, the adder
16a adds the movement results of the multiplexers 15a and 15b,
The adder 16b adds the movement results of the multiplexers 15c and 15d. The addition output of these adders 16a and 16b is the following adder 16c
And output as the density I _R of the converted pixel.
Here, the density of the pixel to be converted is simply 2 representing black or white.
It is assumed that the data is not value data but multi-valued images such as halftone and full-color images, and is represented by 4 bits at each density. In the conversion pixel density calculation circuit 16,
An operation is executed from the result (G1 to G13) of the area determination circuit 14 and the result (p1 to p4) of the high-frequency pattern detection circuit 13 according to an arithmetic operation expression shown in FIG.

FIG. 17 is a block diagram showing a specific configuration example of a multiplexer circuit. The multiplexer 15a (see FIG. 3) is a multiplexer 15a ₁ provided for each digit I _A0 , I _A1 , I _A2 , I _{A3 of} the multi-value density data represented by 4 bits of the input pixel to be converted. 15a ₂ , 15a ₃ , 15a ₄ , based on the presence / absence of high-frequency components from the divided region and the high-frequency pattern detection circuit 13 where the current conversion pixel sent from the divided region determination circuit 14 is located and the pattern information p1 to p4. The multiplexer 15a ₁ , 15a ₂ , 15a ₃ , 15a ₄ comprises a decoder 15e for transmitting information necessary for performing digit shift. The operation of the circuit thus configured is as follows.

(In the case of normal processing) Select G1 3 Select G5, G62 Select G7, G12, G131 Select G8, G9, G10, G110 Select G2, G3, G4 Output 0 (For p1, p2) G1, G4 Output 0 G5, G6, G11, G12 Output 0 Select G2, G3 2 Select GY, G8, G9, G10 2 (For p3, p4) G1, G2 Output 0 G5, G6, G7, G8 Output 0 G3 , G42 is selected. G9, G10, G11, G122 are selected. When G13 is designated by p1 to p4, it depends on the four section areas in FIG.

Here, when the coordinates of the conversion pixel are (X, Y), the thirteenth
In the divided area shown in the figure, the area is divided by the following straight lines. Here, it is assumed that X> 0 and Y> 0. Further, although the curve of the equation (1) is simplified to a straight line, a plurality of pieces of information must be gathered into one conversion pixel according to the conversion magnification. Therefore, the values of the conversion magnifications p and q are used as they are.

If no high-frequency pattern is detected, the following area ratio is taken as usual.

Line pX + qY>0.75; region G3 When the area ratio W _C exceeds 0.75, W _C = 1 and approximate expression is made only with I _C.

Linear pX + qY <0.33; if you do not exceed the area G13 area ratio W _C = 0.33 _{is, W A, W B, W} C, W D
= 1/4 equal combination.

Line 0.33 <pX + qY <0.75; Regions G9, G10 In this case, the second closest pixel is selected by the following equation with W _C = 1/2, and W = 1/4.

Expressed by the region _{G9 W A = 1/8,} W B = 1/4, W C = 1/2, W D = 1/8; (1) linear Y> X.

It is represented by the area _{G10 W A = 1/8,} W B = 1/8, W C = 1/2, W D = 1/4; (2) linear Y <X.

When a high-frequency pattern is detected, W = 1/2 is applied to pixels having high-frequency components (see p1 to p4 in FIG. 8) as shown in FIG.

 Next, the operation of the circuit shown in FIG. 3 will be described.

When calculating the density of the conversion pixel according to the present embodiment,
The converted image is read by the converted pixel input unit such as the transmission unit, and density data of the converted pixel located near the current converted pixel is input. Density data I _A for each of the converted _{pixel, I B, I C, I} D is input to the multiplexer 15a~15d and high frequency pattern detecting circuit 13. Detection of a pixel to be decimated when the pixel is reduced from the position of the conversion pixel and the conversion magnification in the decimated pixel detection circuit 12, and two thresholds or bit planes based on the density data of the input pixel to be converted in the high frequency pattern detection circuit 13. From the determination of the binarization pattern, it is determined whether a high-frequency component appears in the original image and whether or not the information is stored when the image is reduced.

Next, the divided area determination circuit 14 determines the divided area to which the converted pixel belongs based on the position X, Y of the current converted pixel on the converted pixel plane and the conversion magnification p, q. In the present embodiment, as shown in FIG. 13, the division areas are set from the conversion magnifications p and q, and the judgment for each of the division areas is made based on the conditions shown in FIG. FIG.
13 shows a flow of a processing procedure when the divided area determination circuit 14 makes a determination based on the divided areas in FIG. For example, in the drawing, the divided area determination circuit 14 determines whether X ≧ 0, Y
≧ 0, X + Y ≧ 0.75, X + Y ≧ 0.33,
By examining whether or not Y ≧ X, it is possible to determine which divided area the current conversion pixel belongs to.

Note that the sign of X, Y regarding whether X + Y ≧ 0.75, X, Y ≧ 0.33, and Y ≧ X depends on whether X ≧ 0 and Y ≧ 0 or not. In this way, information on the divided area to which the current conversion pixel belongs is sent to the multiplexers 15a to 15d as 4-bit information. A decoder attached to the multiplexer that has received the divided area information (for example, FIG. 17)
15e reference), the partition area and the high-frequency pattern detection conditions necessary to effect the digit movement corresponding to each based on a 2-bit information of each multiplexer 15a ₁ to 15A ₄
And so on. In the multiplexers 15a to 15d and the conversion pixel density calculation circuit 16, in order to obtain the density corresponding to the current conversion pixel from the density data of the pixel to be converted, digit shift is performed based on the divided area determined by the division area determination circuit 14. Or perform an operation.

FIG. 16 shows, for each divided area, when the current conversion pixel is located in the divided area, each density of the pixel to be converted has a corresponding area ratio having a magnitude of one power of two. It describes the contribution of the converted pixel to the density. In other words, each of the area ratios W _A , W _B , W _C , W _{D in} the above-described equation (1)
Is displayed as a power of 2 and the density of each converted pixel located in each divided area is described.

Next, a case where a high-frequency pattern is not detected will be described. If the current conversion pixel belongs to the divided regions G1, G2, G3, and G4 from the divided region determination circuit 12, only the pixels to be converted A, B, and C that are closest to the respective divided regions as shown in FIG. May be considered. Therefore, if the current conversion pixel belongs to the dividing regions G1, the decoder 15e each multiplexer 15a ₁ to 15 multiplexer 15a shown in FIG. 17
a Send a signal to select 3 for ₄ so that no digit shift is performed.

Further, since the current conversion pixel if it belongs to the divided region G5, as is clear from FIG. 16, the contribution to the concentration of the converted pixel from the density I _A of the pixel A is 1/2 of the order, The density data of the pixel A needs to be shifted right by one digit. In addition, as apparent from FIG. 16, in the case of the divided areas G2, G3, and G4, the multiplexer 15a outputs 0 without selecting anything.

When the divided area is G6, the multiplexer 15a selects 2 in FIG. 17 (1/2 contribution).

When the divided areas are G7, G12, and G13, the multiplexer 15a selects 1 in FIG. 17 (1/4 contribution).

When the divided area is G8, G9, G10, G11, the multiplexer 15a
Selects 0 in FIG. 17 (1/8 contribution).

In this way, the densities of the converted pixels shifted by the multiplexers 15a to 15d are added by the adders 16a to 16c of the arithmetic means 16 to obtain the current density of the converted pixels.

Next, when a high-frequency pattern is detected, the following occurs. For example, in the case of FIG. 22, the high-frequency pattern 3 of H1 in FIG. 8 corresponds thereto. Therefore, if the divided area of the conversion pixel is G1, since C / 2 + D / 2 from FIG. 16, 2 in FIG. 17 is selected for the multiplexers 15c and 15d.
Nothing is output to the multiplexers 15a and 15b. In this way, even if a high-frequency component appears in the original image, the information can be stored without being lost at the time of reduction.

[Effects of the Invention] As described above in detail, according to the present invention, even when a high-frequency component appears in an original image, it is not lost during reduction,
While saving information, the area ratio W is set to a power of 2
By doing so, the multiplication of the density I of the pixel to be converted and the W required when calculating the density of the conversion pixel is performed only by bit shift or carry down, and the number of multiplication circuits can be reduced as much as possible. Therefore, according to the present invention, a small-scale circuit and high-speed processing can be realized.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the principle of the first invention, FIG. 2 is a broacher showing the principle of the second invention, FIG. 3 is a block diagram showing an embodiment of the present invention, and FIG. FIG. 5 is an explanatory diagram of thinning data required for detecting a thinned pixel, FIG. 5 is a diagram showing a four-section area, and FIG. 6 is a reference pixel in the case of a reduction ratio 1> p, q ≧ 1/2 according to the embodiment. FIG. 7 is a diagram showing a detection condition of a pixel to be converted to be decimated, FIG. 8 is a diagram showing a high-frequency pattern when a reduction ratio 1> p, q ≧ 1/2 according to the embodiment, and FIG. The figure shows a one-dimensional high-frequency pattern detection, FIG. 10 is a flowchart of a high-frequency pattern detection means, FIG. 11 shows a high-frequency pattern detection by a bit plane, and FIG. 12 shows a high-frequency pattern detection procedure FIG. 13 is a diagram showing a region division according to the embodiment of the present invention, FIG. FIG. 15 is a diagram showing a division condition, FIG. 15 is a flowchart showing a region determination processing procedure of FIG. 13, FIG. 16 is a diagram showing an arithmetic operation expression of the present invention, and FIG. 17 is a block diagram showing a specific configuration example of a multiplexer circuit Fig. 18, Fig. 18 is a block diagram of the principle of the projection method, Fig. 19 is a flowchart showing the processing procedure of the projection method, Fig. 20 is a diagram showing the principle of the projection method, Fig. 21 is an area ratio and logical operation by the projection method FIG. 22 is a block diagram showing a circuit, and FIG. 22 is a diagram showing a case of a 2/3 reduction process by a projection method. In FIG. 1, reference numeral 11 denotes non-converted pixel input means, 12 denotes thinned pixel detecting means, 13 denotes high frequency pattern detecting means, 14 denotes divided area determining means, 15 denotes digit moving means, and 16 denotes converted pixel density calculating means.

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-56-140461 (JP, A) JP-A-63-93269 (JP, A) JP-A-1-185779 (JP, A) JP-A-1- 290373 (JP, A) (58) Field surveyed (Int. Cl. ⁶ , DB name) G06T 3/40 H04N 1/387 101-1/393

Claims (3)

(57) [Claims] 1. A density of each converted pixel is determined based on a projection method from a density of a pixel to be converted in the vicinity of a converted area where the converted pixel is projected on a plane on which the converted pixel is arranged. In a multi-valued image reduction device, a conversion pixel input means for inputting density data of a predetermined conversion pixel group located in the vicinity of a conversion pixel, and detection of original pixels to be thinned out at the time of reduction by the coordinates of the conversion pixels and the conversion magnification A high-frequency pattern detecting unit that detects a high-frequency component by matching a binarized pattern of the pixel to be converted with a high-frequency pattern; Based on the pixel array, each area of the group of pixels to be converted is converted to contribute to the density of the converted pixel by an area ratio having a size corresponding to a power of 2 corresponding to the area within the area. Divided region determining means for determining in which of the divided regions the conversion pixel is located according to the magnification, and the divided region determined by the divided region determining device and the detection data by the high-frequency pattern detecting means A plurality of digit shifting means for shifting the digit of each density data of the pixel to be converted displayed in a binary system based on the digitized data, and converting the converted pixel from the density data of each converted pixel digit shifted by the digit shifting means. And a converted pixel density calculating means for calculating the density of the multi-valued image. 2. The density of each converted pixel is determined based on the projection method from the density of the pixel to be converted in the vicinity of the converted area where the converted pixel is projected on the plane on which the pixel to be converted is located. In the multi-valued image reduction method, a step of sequentially inputting density data of a predetermined group of pixels to be converted located in the vicinity of a conversion pixel (step 1); A step of performing detection (step 2); a step of detecting high-frequency components from density data of the group of pixels to be converted (step 3); Which divided area of the divided area set according to the conversion magnification has the corresponding converted pixel located as one that contributes to the density of the converted pixel by the corresponding area ratio having a size of a power of 2 Judge (Step 4), and the density data of each of the converted pixels displayed in a binary system based on the divided area determined in the step (Step 4) and the high-frequency component detection data in the step (Step 3). A digit shift process (step 5); and a process (step 6) of calculating the density of the converted pixel from the density data of each converted pixel digit shifted in the process (step 5). Value image reduction method. 3. The multi-valued image data high-frequency pattern detecting means includes a binarized pattern (two or more pixels to be converted) based on a plurality of threshold values.
Value image group) and a high-frequency pattern defined in advance, or a bit plane image (binary) is created from multi-value data of each pixel to be converted, and a lower bit plane is converted from a higher bit plane. 3. The multi-value image reducing apparatus according to claim 1, wherein the detection is performed based on a matching condition with a predetermined high-frequency pattern while watching.