JPH04340672A - Method and device for processing image - Google Patents

Abstract

PURPOSE:To obtain the image processing method and device which enables edge emphasis having no transient state of a level by canceling the stepwise change of the density level at the edge part of an input image. CONSTITUTION:For R, G and B chrominance signals read by a scanner, longitudinal and lateral edges are calculated by an edge detection filter, and the respective edges of the respective chrominance signals R, G and B are compared with threshold values T and T' by a comparator 21. According to the compared results, edge directions are decided and when the edge directions are decided, a selector 22 calculates constants A, B, C and D as picture element numbers showing position relation with a considering picture element. Based on level difference between the picture elements designated by those picture element numbers and the considering picture element or the code, the level transforming the considering picture element is outputted through a selector 23.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method and apparatus for detecting and enhancing edge portions of a color image.

0002

[Prior Art] When an image is input using a scanner or the like and the horizontal axis of two-dimensional coordinates is the scanning direction and the vertical axis is the level change, and the level change of the edge part of the input image is observed, the level of the edge part is horizontal. It can be seen that the value does not change sharply after a certain point on the axis, but changes in stages. Conventionally, it has been known to use an edge enhancement filter as a method of enhancing blurred edges. As shown in Figure 10, this involves creating a window around the pixel you want to process (pixel of interest), calculating the sum of the products of the pixels within that window and the weights of the corresponding filters, and applying the results to the pixel of interest. By adding it to the level, it becomes the level after processing.

0003

[Problems to be Solved by the Invention] However, in the conventional edge enhancement method described above, even after processing, the stepwise change in the level of the edge portion does not disappear, and only the beginning and end portions of the edge are affected. I'm just emphasizing it. For example, as shown in Figure 11(a), red on a yellow background,
An image with green text/line drawings is read in using a scanner, and the image represented by the RGB three primary color signals is converted into the so-called L*a*
Convert to b*, take the frequency distribution of a*b* space, and get a*b
*If you look at the distribution, it will look like Figure 11(b). Note that since it is difficult to express the frequency quantitatively here, those with a frequency of 1 or more are represented by black dots.

That is, in the conventional edge enhancement method, as shown in FIG. 11(b), a Milky Way-like connection occurs from the base yellow to red and green. This is because the levels are in a staircase state, so the color changes gradually from yellow to red or green. When edges are emphasized using an edge emphasis filter on the image shown in FIG. 11(a), the result is shown in FIG. 12(a), and the L*a*b* distribution is as shown in FIG. 12(b). It becomes like this. In other words, in the RGB or CMY space, the edges appear to be sharp, but as is clear from Figure 12(b), when viewed in terms of the L*a*b* distribution, the Milky Way-like connection does not appear. Instead of disappearing, extra noise occurs,
The problem is that each color expands.

[0005] The above L*a*b* refers to the hue (
H), brightness (V), and saturation (C), and can be theoretically converted relatively easily as long as it is a color space to which CIE's XYZ coordinate system theory can be applied. Conversion from three primary color signals RGB (NTSC) to CIE-XYZ is as follows: X=0.608R+0.173G+0.200BY=0
．． 299R+0.587G+0.114BZ=
It is performed according to the relational expression such as 0.066G + 1.112B, and L*a is obtained from CIE-XYZ.
Conversion to *b* (CIE) is X0, Y0, Z0
This is executed using the following formula, using the standard color as the X, Y, and Z values.

[0006]

[Math 1]

[0007] Furthermore, hue (H), brightness (V), and saturation (C
) is expressed as follows.

[0008]

[Math 2]

[0009] In image processing, the technology of converting the color itself is known, but the color extraction for this purpose is
B, Extract in CMY 3 primary color signal space, or L*a
Even when extracting in a space that expresses hue, brightness, and saturation, such as *b* and Luv, it is better to have no transient state at the milky-way level mentioned above, both during extraction and color conversion. It is clear that the color areas are clearly separated and easy to process.

The present invention has been made in view of the above, and its purpose is to eliminate stepwise changes in the density level at the edge portions of an input image, and to achieve edge enhancement without level transients. An object of the present invention is to provide a possible image processing method and apparatus.

[0011]

Means for Solving the Problems In order to achieve the above object, the present invention has the following configuration. That is, in an image processing apparatus that reads an image and performs edge processing on the image, there is a means for inputting color signal data of a predetermined number of pixels in the horizontal and vertical directions centering on the pixel of interest, and a means for inputting color signal data of a predetermined number of pixels in the horizontal and vertical directions centering on the pixel of interest, and the pixel of interest is the edge of the image. an edge determining means for determining whether the edge direction is the same; and a means for determining the direction of the edge; a first output means for outputting the converted level of the pixel of interest; and a second output means for outputting the color level of the pixel of interest as is when the edge determining means determines that the pixel of interest is not an edge. Equipped with.

Preferably, the first output means calculates the difference between the color level of the pixel of interest and the color level of a pixel a predetermined number of times before the edge direction of the image, and the difference between the color level of the pixel of interest and the edge direction of the image. On the other hand, when the sign of the difference with the color level of a pixel after a predetermined number of pixels is the same, the pixel of interest is converted based on the difference in color level while decrementing the predetermined number until the predetermined number becomes 1. The subsequent level is output, and when the signs differ, the color level of the pixel with the smaller difference is output.

[0013]

[Operation] The above configuration functions to eliminate stepwise changes in the density level at the edge portions of the input image and to perform edge emphasis without any level transients.

[0014]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings. FIGS. 1 and 2 are block diagrams showing the main components of an image processing apparatus according to an embodiment of the present invention. In FIG. 1, RGB color signals read by a scanner (not shown) are input to a FIFO section 1. This FIFO 1 is composed of five FIFOs, and stores data for five consecutive lines of the image in the vertical direction. Data from the left end to the right end of each line is sequentially output to each latch of the latch group 2 located in the same horizontal direction in synchronization with a clock (not shown). These output data are latched by latch group 2 and stored as 5×5 pixel data in the horizontal and vertical directions centered on the pixel of interest.

The data stored in the latch group 2 is input to the vertical edge calculation section 3 and the horizontal edge calculation section 4, and the vertical edge and the horizontal edge are calculated by the edge detection filter shown in FIG. be done. Among the edge detection filters shown in FIG. 3, (a) is a vertical edge detection filter, and (b) is a horizontal edge detection filter. For example, in the case of a vertical edge detection filter, the method for detecting edges with these filters is as follows: Among the 5×5 data from latch group 2, the pixels corresponding to the weight of -1 and the weight of +1 of the filter are All pixels corresponding to the weights are added individually, and both values after addition are subtracted. With this calculation, 5
A value obtained by applying the vertical edge detection filter to the ×5 data is obtained.

Similar calculations are performed in the horizontal direction, and the edge a (vertical direction) and edge b obtained by these calculations are
(horizontal direction), the color signals R, G,
Find each B.

In the processing circuit shown in FIG.
, G, and B, as described later,
A comparison is made in the comparator 21 with threshold values T and T' set in a register by a CPU (not shown), and the edge direction is determined based on the result. When the edge direction is determined, the selector 22 obtains constants A, B, C, and D, which are pixel numbers indicating the positional relationship with the pixel of interest. R', G'
, B' are output.

The reason why the latch group 2 has a 5×5 configuration in the horizontal and vertical directions is that the applicant has conducted experiments and research on two types of scanners, and found that the level of the color signal changes at the edge portion. This is based on the fact that it was found that a maximum of 5 pixels are required to reach the level.

Therefore, in order to eliminate the above-described transient state of the level of the edge portion, in this embodiment, the circuit shown in FIG. 2 determines whether or not the pixel of interest is an edge portion. If so, calculate the difference in level between the pixel of interest and a pixel before a predetermined number of pixels (m) and a pixel after a predetermined number of pixels in the edge direction, and calculate the difference in level between the pixel of interest and the pixel before and after a predetermined number of pixels in the edge direction. I am trying to force conversion to the pixel level. Note that the number of pixels moved back and forth in the edge direction is m=2, that is, 2 pixels here.

Furthermore, in this embodiment, in order to protect the vertex level of a thin line whose vertex level is only one pixel wide (as shown in FIG. 7), if the signs of the two level differences match, m is further decremented. , m=1, and if the signs of the two level differences match again, the pixel of interest is not converted as a vertex to be saved. However, if the signs of the two level differences do not match, forced conversion is similarly performed to the level of the pixel with the smaller absolute value of the level difference.

Next, the processing procedure in the image processing apparatus according to this embodiment will be explained with reference to the flowchart shown in FIG. In step S1 of FIG. 4, the sum of the absolute values of edge a (vertical direction) and edge b (horizontal direction) for each of R, G, and B is calculated as follows: EDGE=|edge a|+|edge b|
...Find the color signal for which (1) is the maximum, and let it be X. Then, in step S2, it is determined whether EDGE(X) exceeds a predetermined threshold value T, and if it is less than the threshold value, the process proceeds to step S3, in which edge enhancement processing is not performed. The pixel of interest is set to ban=12 with all R, G, and B values unchanged. This ban indicates the positional relationship of pixels to be converted as shown in FIG. 5, and the pixel of interest =
It becomes 12.

On the other hand, if it is determined in step S2 that EDGE(X) exceeds the threshold T, the process proceeds to step S4, where the signs and magnitudes of edge a(X) and edge b(X) are also determined. Find the direction of the edge. Here, if the new threshold value is T'=T/2, |edge a(X)|≧T
If 'and |edge b(X)|<T', the edge direction is assumed to be vertical and the process proceeds to step S5. Also, ｜Etsuji a(X
)|<T' and |edge b(X)|≧T', the direction of the edge is horizontal, and the process advances to step S6.

|edge a(X)|≧T'and|edge b
If (X)|≧T', the direction of the edge is oblique,
Further, calculate edge a(X)×edge b(X), and if the sign is positive, proceed to step S7 assuming that the edge direction is downward to the right; if negative, proceed to step S7, assuming that the edge direction is downward to the left. Proceed to S8.

In the above steps S5 to S8, constants A and B are
, C, and D. These constants A,
For B, C, and D, A is the number of the pixel 2 pixels before the target pixel, B is the number of the pixel 2 pixels after, C is the number of the pixel 1 pixel before, and D is the number of the pixel 1 pixel after. means number. In the following step S9, the difference in level between the pixel of interest and the pixel corresponding to A is determined and set as diff1, and the difference in level between the pixel of interest and the pixel corresponding to B is determined and set as diff1.
Let it be f2. Then, in step S10, diff1×
The sign of diff2 is determined, and if it is positive, there is a possibility that the pixel of interest is the apex of a thin line, so step S12
Proceed to. If the sign is negative, it is assumed that the pixel of interest is in the middle of an edge, and the process proceeds to step S11, where the diff
Based on the absolute values of 1 and diff2, the pixel (A or B) with the smaller value is assigned to ban.

In step S12, the level difference between the pixel of interest and the pixel corresponding to C is determined, and this is set as diff1.
Furthermore, the level difference between the pixel of interest and the pixel corresponding to D is determined and set as diff2. In the next step S13,
From diff1 and diff2 obtained in step S12, d
The sign of if1×diff2 is determined, and if it is positive, the pixel of interest is the apex of the thin line, so the process proceeds to step S14, and ban=12 is set. However, diff1×dif
If the sign of f2 is negative, d is determined in step S15.
The absolute values of if1 and diff2 are determined, and the pixel (C or D) with the smaller value is assigned to ban.

[0026] With the above processing, in all cases b
Since the assignment to an has been completed, that is, the number of the pixel to which the pixel of interest should be converted has been obtained, in step S16, R', G', B' obtained from the horizontal and vertical 5 x 5 data are
The values are R'=R(ban), G'=G(ban), B'
=B (ban). Then, in step S17,
It is determined whether or not processing has been completed for all pixels, and if the result of the determination is YES, the processing is terminated, and if NO, the process returns to step S1 again.

Therefore, for an image as shown in FIG. 6(a), in which when a horizontal edge is cut along a constant plane in the vertical direction, the density level of the edge portion changes stepwise. If the above-mentioned processing is performed, the edge portion will become as shown in Fig. 6 (
In b), it is converted into a smooth edge that changes at right angles, as shown by the solid line. Furthermore, the edges of a thin line image as shown in FIG. 7(a) are improved as shown in FIG. 7(b).

Furthermore, if the image shown in FIG. 11(a) is subjected to the image processing according to this embodiment, the image shown in FIG. 13(a) is obtained.
(a) in Fig. 12 is an image with an edge emphasis filter applied.
Compared to this, the edge portion becomes clearer and the smooth portion becomes a noise-free image. When we examine the L*a*b* distribution of this image, we find that the connection between colors (the Milky Way-shaped part mentioned above) as shown in Figure 13(b) disappears, and the areas of each color become sharp. It can be seen that the distribution is easy to classify into areas.

As explained above, according to this embodiment,
A predetermined image signal is input from the loaded color image around the pixel of interest, and horizontal and vertical edge detection filters are used to determine whether the pixel of interest is an edge portion.
If it is an edge, find the direction of the edge and find the difference between the level of the pixel of interest and the level of a predetermined pixel in the edge direction.If the pixel of interest is not an edge, output the level of the pixel of interest as is. This has the effect that the stepped edge, which is a transient state of the level of the edge portion of the image, is improved to an edge shown by a straight line or right angle, and the color boundary becomes clear. Furthermore, since noise is not emphasized even though edges are emphasized, there is an effect that regions of each color existing in an image in each color space can be easily extracted.

<Modification> In the above embodiment, the edge detection filter shown in FIG. 3 was used in edge detection to determine whether the pixel of interest is an edge.
Edge detection may be performed using a 3×3 filter shown in ).

When this filter is used, the number of FIFOs for inputting RGB color signals is reduced to three, and the number of latches is also reduced to approximately 1/3, thereby simplifying the processing circuit. Furthermore, in the above embodiment, the edge direction is divided into four directions, but if these are limited to two directions, the vertical direction and the horizontal direction, the filter required to detect the edge will be a simple one as shown in FIG. Therefore, the edge detection process itself can be simplified.

Furthermore, if the scanner has a resolution of 400 dpi or more, the presence of thin lines as shown in FIG. 7 can be ignored, and steps S12, S13, and S15 in the flowchart shown in FIG. 4 can be omitted. On the other hand, when performing edge detection on a scanner with a resolution of 200 dpi or less, the number m of pixels before and after the pixel of interest, which is necessary to determine the level difference with the pixel of interest, can be calculated by only processing m = 1. This is sufficient, and the processing of steps S9 to S11 in the flowchart shown in FIG. 4 can be omitted.

Furthermore, although the above embodiment has described edge detection processing for color images, it goes without saying that this processing can also be applied to black and white images. Note that even if the present invention is applied to a system composed of a plurality of devices, only one
It may also be applied to a device consisting of two pieces of equipment. It goes without saying that the present invention can also be applied to cases where the present invention is achieved by supplying a program to a system or device.

[0034]

[Effects of the Invention] As explained above, according to the present invention,
By determining the edge direction of the input image and calculating the converted level of the pixel of interest based on the level difference between the level of a predetermined pixel along that direction and the pixel of interest, the density level level of the edge portion of the image can be determined. This has the effect of eliminating state changes and realizing edge enhancement without level transients.

[Brief explanation of drawings]

[Figure 1]

FIG. 2 is a block diagram showing the main part configuration of an image processing device that is an embodiment of the present invention;

FIG. 3 is a diagram showing an edge detection filter according to an embodiment;

[
FIG. 4 is a flowchart illustrating the processing procedure in the image processing device according to the embodiment;

FIG. 5 is a diagram showing the positional relationship of pixels to be converted;

[Figure 6]
An image in which the density level of the edge portion changes stepwise, and a diagram showing the edge portion after processing,

FIG. 7 is a diagram showing a thin line image and an edge part after processing,

[Figure 8] 3 according to modification example
A diagram showing a ×3 filter,

FIG. 9 is a diagram showing a filter for detecting an edge direction according to a modification;

FIG. 10 is a diagram showing a conventional edge emphasis filter;

[Figure 1
1] A diagram showing an image represented by RGB three primary color signals read by a scanner and its frequency distribution in a*b* space,

FIG. 12 is a diagram showing an image obtained by applying edge enhancement to the image of FIG. 11 using a conventional edge enhancement filter, and its L*a*b* distribution;

FIG. 13 is a diagram showing an image after processing the image in FIG. 11 and the L*a*b* distribution of the image.

[Explanation of symbols]

1 FIFO 2 Latch group 3 Vertical edge calculation section 4 Lateral edge calculation section

Claims (3)

[Claims] 1. An image processing device that reads an image and performs edge processing on the image, comprising: means for inputting color signal data of a predetermined number of pixels in the horizontal and vertical directions centering on a pixel of interest; Edge determination means for determining whether a pixel of interest is an edge of an image; means for determining the direction of the edge; and a color level of the pixel of interest and a color level of a pixel shifted forward or backward by a predetermined number with respect to the edge direction of the image. a first output means that outputs the converted level of the pixel of interest based on the difference between the pixel of interest and the edge determination means that determines the color level of the pixel of interest when determining that the pixel of interest is not an edge; An image processing apparatus comprising: a second output means for outputting the image as it is. 2. The first output means outputs the difference between the color level of the pixel of interest and the color level of a pixel a predetermined number of times earlier with respect to the edge direction of the image, and the difference between the color level of the pixel of interest and the edge direction of the image. When the difference between the pixel and the color level of the pixel after a predetermined number of pixels has the same sign, the predetermined number is decremented until the predetermined number becomes 1, and the pixel of interest is converted based on the difference in color level. 2. The image processing apparatus according to claim 1, wherein when the signs are different, the color level of the pixel with the smaller difference is output. 3. An image processing method that reads an image and performs edge processing on the image, comprising: inputting color signal data of a predetermined number of pixels in horizontal and vertical directions centering on a pixel of interest; A step of determining whether the pixel of interest is an edge of the image, a step of determining the direction of the edge, and a difference between the color level of the pixel of interest and the color level of a pixel shifted forward or backward by a predetermined number with respect to the edge direction of the image. Based on the step of outputting the converted level of the pixel of interest and the step of determining the edge, when it is determined that the pixel of interest is not an edge, the color level of the pixel of interest is output as is. An image processing method comprising the steps of: