JPH01213761A - Edge detection system and transmission processing system for color image - Google Patents

Abstract

PURPOSE:To directly detect the edge of a color image from a chromaticity component by comparing the divergence component output and the rotation component output with a prescribed reference value. CONSTITUTION:A separation processing part 10 consists of a luminance/ chromaticity separating part 11 and an operation processing part 12 which separates color components to a lamellar component and a vortex component. A transmission processing part 20 consists of a lamellar transmission processing part 21, a vortex transmission processing part 22, and a contour line processing part 25. An edge extracting part 26 compares the divergence component output of a lamellar vector analysis connection part 23a and the rotation component output of a vortex vector analysis connection part 23b with a prescribed reference value to detect the edge of the color image.

Description

[Detailed Description of the Invention] [Summary] A vector analysis of differential action is performed on the lamella component and the vortex component, which are obtained as a result of applying Helmholtz's theorem to the chromaticity component in a color image.7 The output value obtained by this vector analysis is This invention discloses that edges of a color image can be detected based on the size of the vector, and that color images can be synthesized based on the size of the output value determined by this vector analysis.

[Industrial application field]

The present invention relates to an edge detection method and a transmission processing method for color images, and in particular, the present invention separates chromaticity components into lamellar components (laminar flow components) and vortex components (vortex components), and generates color images from the separated chromaticity components. The present invention relates to a color image edge detection method and a transmission processing method that make it possible to detect the edges of a color image and realize efficient transmission of a color image using the edge information thus detected. .

[Conventional technology]

Conventionally, edges in a color image have been detected by solving the structural line equation expressed by the following equation. Here, from the C-line there is a ridge line/valley type edge, from the D-line there is a dividing line type edge,
There are step-shaped edges from the E-line and L-line.
It will be sought after. Note that the scalar function Φ<x, y>
In the final step, the brightness information of the color image is used.

C (x, y) = XゝHX, Tomi O: C-line D (x,
y) =Xi HXJL =0: D-line E(x, y)=XtHX=0: E-line L(x, y) =trace H=0
: L-line however.

Furthermore, conventionally, when transmitting and reproducing color images,
As shown in the case of television transmission, luminance components and chromaticity components are transmitted corresponding to pixels on a color image. And when the three primary colors are R, G, and B.

(i) As brightness.

L = 0.2988R + 0.5868G + 0.114
4B(II) Chromaticity vector I, Q as 1 =0.736(R-L) -0,286(B-L
)Q =0.478(R-L) +0.413(B
-L) are used to transmit the values I and Q given respectively. That is, the above 1. For example, Q is placed on a subcarrier and transmitted.

[Problem to be solved by the invention]

However, in the conventional edge detection method for color images, edges are detected using luminance information, so even if an edge exists, there is no significant difference in luminance. A major problem was that edges could not be detected.

Furthermore, although color images are conventionally transmitted in the manner described above,
The chromaticity signal (signal regarding the chromaticity component) of a color image has a considerable energy component. Therefore, if it is possible to reduce part of the energy of the chromaticity signal, it is possible to compress the transmission band or This is advantageous when processing information regarding the three chromaticities.

The present invention has been made in view of such circumstances, and
We have previously filed a patent application as Japanese Patent Application No. 62-133690, and in connection with this patent application, we are attempting to provide a color image edge detection method that allows the edges of color images to be detected from chromaticity components. It is something to do. Furthermore, in connection with the above-mentioned patent application, the present invention aims to provide a color image transmission processing method that allows the chromaticity signal band of a color image to be widened (compressed).

[Means to solve the problem]

FIG. 1 is a diagram showing the principle configuration of the present invention.

In the figure, 10 is a separation processing unit, which includes a luminance/chromaticity separation unit 11 that separates the luminance component and chromaticity component of pixels forming a color image; The chromaticity component is divided into a lamella component and a vortex component by applying Helmholtz's theorem. 20 is a transmission processing section.

It is composed of a lamella transmission processing section 21, a vortex transmission processing section 22, and a contour processing section 25. The lamella transmission processing unit 21 includes a lamella vector analysis connection unit 23a that outputs a divergence component that expresses the lamella components separated by the arithmetic processing unit 12 as a differential, and converts the divergence component into an edge.
It is composed of a divergence component output of the lamella vector analysis connection section 23a and a lamella encoding processing section 24a that band-compresses and codes the lamella components separated by the arithmetic processing section 12. The vortex transmission processing section 22 includes a vortex vector analysis connection section 23b that converts the output representing the vortex components separated by the calculation processing section 12 as a differential into an edge, and a rotation he output and calculation of the vortex vector analysis connection section 23b. The vortex coding processing section 24b band-compresses and codes the vortex components separated by the processing section 12.

The contour processing unit 25 includes a lamella vector analysis connection unit 23
an edge extraction section 26 that detects edges of a color image by comparing the output value output from either or both of the vortex vector analysis connection section 23a and the vortex vector analysis connection section 23b with a predetermined reference value; and the luminance/chromaticity separation section 11. The brightness edge extraction section 27 detects the edges of a color image based on the brightness components separated by the brightness edge extraction section 27. Reference numeral 30 denotes a composition processing section, which receives the code information from the lamella encoding processing section 24a and the vortex encoding processing section 24b and the contour information from the 2 contour processing section 25 and composites a color image. .

(Function) According to the present invention, the divergence component output that expresses the lamella component outputted by the lamella vector analysis connection unit 23a as a differential, and the vortex vector analysis connection unit 2
The rotational component output, which expresses the vortex component outputted by 3b as a differential, takes a large non-zero value near the edge of the color image, which is caused by the difference in nine chromaticities, unlike in areas other than the edge. from now.

The edge extraction unit 26 is connected to the lamella vector analysis connection unit 23.
Edges in a color image can be detected by comparing the divergence component output of a and the rotation component output of the vortex vector analysis connection section 23b with a predetermined reference value. Therefore, edges with a difference in chromaticity can be detected even if there is no difference in luminance itself.

In this color image edge detection method.

By adding the edges detected by the luminance edge extraction section 27, edge detection of color images becomes more reliable.

According to the present invention, the lamella component and vortex component on a specific axis and the lamella vector analysis connection section 23
If the divergence component output of the lamella component outputted by a and the rotation component output of the vortex component outputted by the vortex/vector analysis connecting section 23b are encoded and sent to the synthesis processing section 30, everything can be done on the 2 synthesis processing section 30 side. This means that the chromaticity components on the axis can be synthesized. Moreover, at this time, the divergence component output and the rotation component output can be considered to be almost zero outside of the edges, and the band compression method for monochrome images can be effectively applied. Therefore, color images can be transmitted by encoding very little information.

〔Example〕

Hereinafter, the present invention will be explained in detail according to examples.

FIG. 2 shows a configuration diagram of an embodiment of the present invention. In the figure.

The same symbols are used for the same components as in FIG.

As explained in FIG. 1, the luminance/chromaticity separation unit 11
Luminance component 13 and chromaticity component 14 of a pixel of a color image
The arithmetic processing unit 12 separates the nine chromaticity components 1
4 is separated into a lamellar component 15 and a vortex component 16. Here, we have a vector V with 9 chromaticity bases and 14.

preferably instead of ξ and η.

It is better to perform weighting based on brightness and use chromaticity components ξ' and η' for weighting such as 5th order.

ξ′ =ξ/ (a+ (1-a)L)η′ =v/
It is preferable to use (a+ (1-a)L). However, a is used to apply a bias so that the chromaticity component vector becomes uniform between O and 2π, and is given by O<a<1.

As also explained in FIG. 1, the lamella transmission processing section 21
is composed of a lamella vector analysis connection section 23a and a lamella encoding processing section 24a, and the vortex' transmission processing section 22 includes a vortex vector analysis connection section 23a.
The contour processing section 25 is composed of an edge detection section 26 and a luminance edge extraction section 27. The contour processing unit 25, which detects the edges of the color image, uses the contour information of the color image formed by the edges (if the contour is approximated by a straight line vector, the position and length of the vector, the starting point of the vector)・The lamella encoding processing unit 24a and the vortex encoding processing unit 24b
This provides position information for alignment during encoding. Conversely, the arrows pointing from the lamella transmission processing section 21 and the vortex transmission processing section 22 to the contour processing section 25 indicate that the edge extraction section 26 is
This is because the edges of the color image are detected from the divergence component output of the lamella vector analysis connection section 23a and the rotation component output of the vortex vector analysis connection section 23b.

31 is a lamella synthesis section, which is for synthesizing the code signals sent from the lamella transmission processing section 21; 32 is a vortex synthesis section, which is used to synthesize the code signals sent from the lamella transmission processing section 21;
For synthesizing the code signals sent from 2,
33 is an interpolation processing unit for performing interpolation processing on the luminance component transmitted via the 1-contour processing unit 25; 34 is an image reproduction unit for processing the transmitted luminance component and lamella; It is for reproducing a color image based on the component and the vortex component. The lamella synthesis section 31, the vortex synthesis section 32, the interpolation processing section 33, and the image reproduction section 34 constitute a synthesis processing section 30.

The reason why the interpolation processing unit 33 is provided here is that the transmitted luminance component may be the luminance of all pixels on a color image, but generally the luminance of pixels at discrete positions is transmitted. This is because the brightness of the pixel located at the intermediate position between the two pixels is determined by interpolation from the brightness of the pixel at the discrete position. Furthermore, even when transmitting the luminance of all pixels, interpolation may be performed, for example, to recover luminance information lost during transmission.

Note that the interpolation processing section 33 and the lamella synthesis section 31 are mutually connected.

Then, one position information is exchanged between the interpolation processing section 33 and the vortex synthesis section 32.

Next, the separation process executed by the arithmetic processing unit 12 will be explained.

A vector of chromaticity component 14.

If it is expressed as, then applying Helmholtz's theorem to this vector ■.

V = grad L + rot(R-K) −
+11 Formula where: L(x, y): Scaling potential such as luminance R-K: Vector potential whose direction is indicated by the unit vector K in the Z-axis direction ai i-x, y R Ri = - ai i x, y+1) Let us calculate div V (hereinafter sometimes referred to as divergence processing) for the vector ■ of the equation.

div V= div-gradL + div-ro
t(R-K)” div・gradL.From now on.

ξx + ηy= Lxx+ Lyy −(2)
A formula is derived. Then, rot V is calculated (hereinafter sometimes referred to as rotation processing) for the vector ■ in equation (1).

rotV=roigradL+rot-rot
(R-K)w rot-rot(R-K). from now.

ξy 77 x = Rxx + Ryy −
(Equation 31 is derived. However, fx, fxy...
etc., partial differential af/ax, θ”f/3x ay・
It represents... In the above equations (2) and 0), each left side can be measured, and L and R can be solved by solving equations (2) and (3).

In this way, gradL and ro can be found by applying Helmholtz's theorem to the vector ■ of one chromaticity component 14.
In the present invention, the two t(R-K) will be referred to as a lamella component 15 and a vortex component 16, respectively.

In FIG. 3, lamella component 1 determined by the arithmetic processing unit 12
5 and 0 illustrating an example of calculation of the vortex component i6, 4 is a given color image, and represents that a disk portion 1 of a different color is placed in a background portion 2 of a different color. A vector line 3 represents the chromaticity components of such a color image. The straight line portion of the vector line 3 extending in the illustrated horizontal direction represents that the background color of the background portion 2 is given. Also, the straight line portion of vector line 3 extending in the diagonal direction shown in the figure is 9
This indicates that the disk portion 1 is given a color different from the background color. The lamella component 15 and vortex component 16 found from such a color image are expressed as 5 in Figure 9.
Shown below. As shown in this figure, the lamellar component 15 represents a laminar flow component, and the vortex component 16 represents a swirl bone.

In the above, grad L and rot(R-K) may be separated into a lamellar component and a vortex component, respectively, or the lamellar component and the vortex component may be differentiated by their differentiation.

In order to express the above differential div V, the differential expression of the lamellar component and the differential expression of the vortex component can be obtained by vector signals div V and rot V of one chromaticity, respectively. The arithmetic processing unit 12 performs arithmetic processing for this purpose. In this case, the vector analysis connections for each component may be div V and rot V.

Next, a description will be given of how the present invention configured as described above can detect edges of a color image. As shown in Figure 4.

Area A and area B have different chromaticity vectors.

It is assumed that they are continuously connected via a C region having a width of 2a. Here, the chromaticity vector ■ of the A area. , B.S.
chromaticity vector v3 in U area, respectively.

Assume that In such an edge part,
If you try to find div V and rot V, di
From the definition of v■.

θx 9y’la = − (rb cos β−r, cos α) a is derived from the definition of rot V.

ay θx 2a = − (r% sin β − ra sin α) a will be guided.

As is clear from these equations, 1divVl and 1
rotVI shows a large non-zero value near the edge of a color image, and also shows a value that is almost zero outside the vicinity of the edge.

Therefore, if one or both of IdivVl and 1rotV1 is detected to have a value larger than a predetermined value, edges of a color image can be extracted. Here, as is clear from equation (2), 1divVl can be found as component 15 by performing a direct divergence operation on the 9 chromaticity body number V and using the lamella component as a differential value, and 1rotVl is (3 ) As is clear from the equation, it can be found by directly performing a rotation operation on the nine chromaticity body numbers (■). From this, the edge extraction section 26 compares the divergence component output of the lamella vector analysis connection section 23a and the rotation component output of the vortex vector analysis connection section 23b with a predetermined reference value set near zero. This means that the edges of the image can be detected.

When the edge detection method of the present invention configured as described above is used in combination with the conventional edge detection method based on brightness, edge detection of color images becomes more reliable. With such a combination, edges that have no significant difference in luminance but a significant difference in chromaticity and edges that have no significant difference in chromaticity but a significant difference in luminance can be distinguished. This is because it becomes possible to detect everything without exception.

Next, the color image transmission processing method of the present invention will be explained.0 The reason why the present invention separates the chromaticity component 14 into the lamella component 15 and the vortex component 16 is basically to reduce the energy of the chromaticity signal by a small amount. However, by making it smaller, we were able to compress the transmission band.

However, since there is a property that 1divVl and IrotVl exhibit values that are almost zero except near the edges of a color image, it becomes possible to use this property to further compress the transmission band. This transmission processing method will be explained below.

At the sending end, the chromaticity component 1 along a particular X-axis of the color image
The value of the vector (ξ, η) of 4 (i.e.

lamella component and vortex component) are prepared, (ξX, ηX) is (ξ, ηX) between adjacent pixels.
Since it corresponds to the difference value of η), it can be easily obtained. Therefore, if the values of div V and rot V are given, according to equations (2) and (3) above,
(ξy, ηy) can also be easily determined.

From this, (ξ) at that particular X-axis.

By solving the difference equation of (ξy, yy) with the value of I) as the initial value, (ξ, η
) can be reproduced, and by repeating similar processing, it is possible to reproduce (ξ, η) over the entire area of the color image.

That is, the receiver receives the values of div V and rot V.

Extremely small lamella component 15 and vortex component 1
By taking a value of 6, the chromaticity component 14 of the color image can be reproduced. Moreover, diν
■ and rot V are, so to speak, monochrome images, and the values are zero in most areas of the color image, so
If a monochrome image is encoded and transmitted using a band compression method (for example, gradient orthogonal transform, cosine transform, or selective DPCM method), accurate color images can be reproduced with an extremely small amount of transmission. You will be able to do it.

FIG. 5 shows a specific example of edges extracted from a color image using the color image edge detection method of the present invention. The edge detection method for this data relies on the luminance component (an edge detection method is also used. Figure 5 shows the original output of the image displayed on the CRT display as a hard copy. Comparison with the original color image confirmed that the edge detection method of the present invention was able to detect edges more accurately than the conventional method.

〔Effect of the invention〕

As explained above, according to the present invention, edges of a color image can be directly detected from the chromaticity components. Therefore, even edges with no significant difference in brightness can be detected accurately.

Furthermore, according to the present invention, it is possible to reproduce a color image using encoded information of a significantly smaller number of chromaticity components.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the principle configuration of the present invention. FIG. 2 is a configuration diagram of an embodiment of the present invention. Figure 3 shows an example of separation of the lamella/vortex separation section. FIG. 4 is an explanatory diagram for edge detection. FIG. 5 shows a specific example of edge detection. In the figure, 10 is a separation processing section, and 11 is a luminance/chromaticity separation section. 12 is an arithmetic processing unit, 20 is a transmission processing unit, 21 is a lamella transmission processing unit, 22 is a vortex transmission processing unit, 23a
23b is a lamella vector analysis connection section, 24a is a lamella coding processing section, and 24b is a vortex coding processing section. 25 is a contour processing section, 26 is an edge extraction section, 27 is a brightness edge extraction section, 30 is a synthesis processing section, 31 is a lamella synthesis section, 32 is a vortex synthesis section, and 33 is an interpolation processing section. 34 is an image reproduction section.

Claims (4)

[Claims] (1) In a color image edge detection method for detecting edges of a color image formed from pixels having a luminance component and a chromaticity component, the chromaticity component of the pixels forming the color image is treated as a lamellar component. an arithmetic processing section (12) for separating the lamella components into vortex components; and a lamella/vector analysis connection section (23a) that outputs a divergence component that expresses the lamella component separated by this arithmetic processing section (12) as a differential and converts it into an edge. ), a vortex/vector analysis connection unit (23b) that outputs a rotational component that expresses the vortex component separated by the arithmetic processing unit (12) as a differential, and a vortex/vector analysis connection unit (23b) and the lamella. - An edge extraction unit (26) that detects edges of a color image by comparing the output value of one or both of the vector analysis connection units (23a) with a predetermined reference value; Features a color image edge detection method. (2) The color image edge detection method described in item 1 further includes a brightness edge extraction unit (27) that detects edges of the color image based on brightness components of pixels forming the color image, An edge detection method for a color image, characterized in that an edge of a color image is detected by an edge detected by an edge extraction section (27) and an edge detected by an edge extraction section (26). (3) A color image is decomposed into pixels, and regarding the luminance component and chromaticity component corresponding to the decomposed pixel, the vector signal making up the chromaticity component is transmitted in correspondence with the scalar signal corresponding to the luminance component. In addition, in a color image transmission processing method for reproducing the color image based on the transmitted scalar signal and vector signal, for a given color image, a luminance component and a chromaticity component are determined corresponding to the pixel. a separation processing unit (10) that further separates the separated chromaticity components into lamellar components and vortex components; and a divergence processing unit (10) that expresses the lamellar components separated by this separation processing unit (10) as a differential. A vector analysis connection unit (23a, b) that outputs the components and outputs an edged rotation component that expresses the separated vortex component as a differential; and an output obtained by this vector analysis connection unit (23a, b). a coding processing unit (10) that codes the value using a monochrome image band compression method, and also codes differential values and boundary values for the lamella components and vortex components separated by the separation processing unit (10); 24a,b
), a contour processing section (25) that extracts a contour of a color image from the output value of the vector analysis connection section (23a, b) and the luminance component of the pixel, and the encoding processing section (24a, b) encoded information;
a composition processing section (30) for synthesizing the color image in response to the transmission of the contour information from the contour processing section (25); The above chromaticity component is used as a reference value, and the chromaticity component of the entire area of the color image is calculated from this reference value using the above-mentioned divergence component output and the above-mentioned rotation component output. A color image transmission processing method. (4) In the color image transmission processing method described in item 3, edge detection is performed using a luminance signal and a chromaticity signal,
A color image transmission processing method that divides the signal image into several segments, encodes the luminance signal and chromaticity signal independently within each segment, and transmits the signals.