JPS5952768B2 - Color figure feature extraction method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method for identifying various colored figures having texture.

An example of a colored figure with texture is shown in a and b of FIG.

In the figure, the Red part is a red part, and the Green part is a green part.

The color figure shown in Figure 1a is a green background with red stripes running from the upper right to the lower left, and Figure 1b is a color figure with the stripes running from the upper left to the lower right. It is.
When identifying these two color figures, the conventionally well-used method is to first scan each point of the color figure with a color television camera to obtain three primary color signals at each scanning point; , g, b are introduced into a predetermined function C2f(ni, g, b), the C value at that time is obtained as a color code, the appearance frequency of this is calculated, and the appearance frequency is used to create a color figure. It is used as a means of identification. in this case,
If the color code of each point in the red area (Red) in Figure 1 a and b is R, and the color code of the green area is G, then in Figure 1 a and b, the color codes R and G appear on the entire screen. The frequencies will be the same and no difference will appear. In the case of Figure 1 a and b
It is assumed that the red regions Red have the same area.

The same applies to the green area Green. ) As described above, it has been difficult to identify colored figures with textures such as those shown in a and b of FIG. 1 using conventional point-by-point processing.

In other words, the identification accuracy was low. An object of the present invention is to provide a color graphic feature extraction method with high identification accuracy.

In order to achieve the above object, the present invention provides a color code group of scanning points existing within a certain spread range centered on a certain scanning point, that is, a set of a certain fixed number of scanning points. Detects the characteristics (patterns) of
The number of color code groups showing the same color code distribution (pattern) within a predetermined observation area is counted to obtain the frequency of appearance of various color code groups, and this frequency is used as a color figure identification means.

In other words, the present invention introduces the concept of microtexture, which takes into account the color code in the vicinity of a scanning point as a parameter associated with each scanning point, and calculates the frequency of appearance of microtextures with the same characteristics. By calculating this, we obtain the characteristic quantities for identifying colored figures with texture.

An embodiment of the present invention will be described below.

FIG. 2 shows an example of how to create color code groups. The coordinates are determined as shown in the figure, and the color code of the point (X, y) is C(X
.
x+1, y), C(X, y+1), C(X+1, y+1
). The color code quartet C (X,
y), C(x+1, y), C(X, y+1), C(X+
1,y+1) is defined as the microtexture at point (X,y). If the microtexture of each point of the color figure shown in FIG. 1A is determined, a microtexture pattern as shown in FIG. 1C will appear for the points of the entire color figure shown in FIG. That is, the pattern C-1 appears in the red region Red, or the pattern C-2 appears in the green region Green.

At the boundary between the red area Red and the green area Green, C-
Patterns 3 to C-6 will appear. Figure 1b
The appearance pattern of microtexture for the color figure shown in FIG. 1d is shown in FIG. As is clear from comparing the two, the difference in the direction of the striped pattern in the red area in Figure 1 a and b is the difference between the pattern in Figure 1 C (7) C-3 to C-6 and the pattern in Figure 1 d <
2) It will appear as a difference from the patterns d-3 to d-6. That is, when calculating the appearance frequency of mital texture in the color figures shown in Fig. 1A and b, C-1 and C
Patterns other than 2, d-1, and d-2 are patterns C-3 to C-6 in the color figure in Figure a, and patterns d-3 to d-6 in the color figure in Figure 1B. appear dominant,
By comparing these patterns, it becomes possible to identify the colored figures A and b. In this identification, if any of the patterns described above do not match, the pattern will be identified as a different colored figure. Although Fig. 1 describes the case of identifying the difference in the direction of the stripes, Fig. 4 shows a scale-connected pattern consisting of a repeating scale pattern.

In the same figure, the case where the scale pattern is upward is shown in Fig. 4a, the case where it is downward is shown in Fig. 4b, the microtexture pattern of Fig. 4a is shown in Fig. 4C, and the pattern of Fig. 4b is shown. The patterns of the Mitaro texture are shown in FIG. 4d, and in each figure, as in the case of FIG.
The color drawing, which is a pattern, can be identified depending on whether or not the numbers .about.f-8 match each other. Note that the color codes of the triangular areas (an enlarged view of which is shown in FIG. 4e) divided by diagonal lines in the microtextures shown in FIGS. 4C and d are formed by dividing the microtextures into multiple equal parts. The second microtexture is detected in the same manner as above and serves as the basis for creating the pattern. Therefore, the pattern in that part is identified based on the predominant frequency of this pattern. In addition, in the case of character discrimination, as shown in FIG. 5, the discrimination between the forward facing A shown in FIG. 5a and the reverse facing A shown in FIG. can.

Furthermore, in the above embodiment, a single color pattern has been described, but even if the color tone is different, the color code can be read out corresponding to the color, and the color pattern can be identified in the same way as described above.

FIG. 3 shows a block diagram of an example of the configuration of an apparatus for carrying out the present invention.

This will be explained below. Reference numeral 1 denotes a color figure input device consisting of a color television camera or the like, which sequentially scans the target color figure and outputs three primary color signals R.1 at each scanning point. g
．． Output b.

2 is the three primary color signal R. g. This is a color analysis circuit that performs conversion C=f (R.g.b) from b to N types of color codes C.

As the transformation f, for example, R, <r<r',, G, <g<
1 and Bi<b<b″, then Ci=f(R.g.
b) ((R.g.b) are the three primary color values of the point (X, y),
Let the color code of point (X, y) be C. However, r1, r″
1, G,, g',, b1, b''1, C,, (1=1~
N) is a predetermined value. ) can be considered. 3 is a color code memory that stores color codes over the entire screen, and its contents are read out by a readout circuit 4.
Read by.

The readout circuit 4 can simultaneously read out the color codes of 2×2 points in the vicinity of the designated coordinate point. Reference numeral 7 denotes a coordinate selection circuit that sequentially reads M coordinate values stored in the address register 8 and sends them to the reading circuit 4.

5 is a determination circuit that determines the characteristics of the 2×2 code pattern read out by the readout circuit 4, that is, the color code group, and generates a corresponding address signal according to a predetermined rule. .

6 is a counting circuit having N4 counters, each of which corresponds to each expected microtexture pattern, and calculates the frequency of appearance of each pattern.

Next, the operation of this device will be explained.

The target color figure is read by the color figure input device 1, and each point of the color figure is input to three original signals R. g. b and is output to the color analysis circuit 2. Read signal R.
g. b is converted into a color code C by the color analysis circuit 2 according to a predetermined conversion rule, and this color code C is stored in the corresponding coordinate portion of the color code memo 1J3. Now, if the coordinates of the point specified by the coordinate selection circuit 7 are (X, y), then C(X, y
), C(x+1,y), C(X,y+1), C(X+1
, y+1) are read out through the readout circuit 4.

The judgment circuit 5 uses this color code pattern (there are N4 types)
and outputs an address signal according to established rules. (For example, the color code is represented by numbers 1 to N,
Four color codes C(X, y), C(X+1, y), C(
For (X, y+1) and C(X+1, y+1), (C(
X, y)-1・N3+(C(x+1,y)-1)・N2
+(C(X,y+1)1)・N+C(X+1,y+1)
may be used as the address signal. ) The counting circuit 6 increments the counter of the address corresponding to the input address signal by one. When the addition operation in the counting circuit 6 is completed, a completion signal is transmitted to the coordinate selection circuit 7 through the signal line 9, and the circuit 7 reads out the next coordinate value from the address register 8, transfers it to the readout circuit 4, and performs the following operations. Same operation as address register 8
Repeat until there are no more contents. As a result, the appearance frequency of the same microtexture in the specified coordinate point group of the color figure is obtained. Therefore, by comparing the output value of the counting circuit 6 with the reference value after scanning one color figure, it is possible to know whether the color figure is different from the reference color figure represented by the reference value. .

As described above, in the present invention, a color figure is determined based on the frequency of appearance of various color code patterns in an area consisting of a set of a plurality of scanning points, rather than the frequency of appearance of each color code at each scanning point of a color figure to be observed. Since the identification accuracy is improved,
It becomes possible to identify colored figures, which was difficult in the past.

[Brief explanation of the drawing]

Figures 1A and b are plan views showing color figures, and Figure 1C
, d are partial views of each part. Fig. 2 is an explanatory diagram showing the coordinates representing each part address of a color figure, Fig. 3 is a block diagram showing an example configuration of an apparatus for carrying out the present invention, and Figs. 4A and 4h show color figures of the picture. FIG. 4C and d are plan views of each pattern, FIG. 4E is an enlarged view of the second microtexture that makes up the pattern, and FIGS. 5A and B are plan views showing the color figures of letters. be. 1・
... Color graphic input device, 2... Color analysis circuit, 3...
- Color code memory, 4... Readout circuit, 5... Judgment circuit, 6... Counting circuit, 7... Coordinate selection circuit, 8
...Address register, 9...Completion signal line.

Claims (1)

[Claims] 1. A color figure to be observed is scanned by a color figure input device, and the three primary color signals at each scanned point are converted into a specific color code by a color analysis circuit and stored. , various color code groups consisting of color codes of scanning points existing in a range including that point and a plurality of points in the vicinity of this point are obtained as patterns by a color code reading circuit from a storage device, and the above color code groups are obtained. A color figure feature extraction method characterized by identifying a color figure based on whether or not a plurality of patterns created based on the same match.