JPH077739A - Color identifying and extracting device - Google Patents

Abstract

PURPOSE:To set a color intended by the user, to identify and extract the color and to make the device to be hardly affected by a change in the lightness of an input picture by providing a mapping means, a distance arithmetic operation means and a detection output means. CONSTITUTION:ROM tables 5, 6, 7 being a map means map an input color video siganl in a feature space with features relating to a color as respective axes. Distance arithmetic operation ROM tables 11, 12, 13, being a distance arithmetic operation means obtain the distance between a color set by a color setting means on the mapped characteristic space and a color represented by an input color video signal. Magnitude comparators 14, 15, 16 and an AND gate 17 being a detection output means output a detection signal when the obtained distance is a preset distance or below. Then the input color video signal is converted into a feature space having high independency and linearity with respect to a human sense to execute color identification and extraction, then the color is stable sgainst a change in the lightness and well matched with the human sense.

Description

Detailed Description of the Invention

[0001]

The present invention is used in image processing. Particularly, the present invention relates to a color identification extraction device that identifies and extracts a certain color from a color image.

[0002]

2. Description of the Related Art In order to identify and extract a certain color from a color image picked up by an image pickup device such as a color camera, conventionally, a window comparator or the like is provided for each of RY and BY input signals. I was extracting a certain range. That is, the upper limit value and the lower limit value are set for each of the R-Y and B-Y input signals, and the signal within the range is extracted as a desired color.

[0003]

However, the upper limit value and the lower limit value of the window comparator to be set for each of the RY and BY input signals are determined by the user while observing the extraction result. There is a need to. RY
The input signals B and B respectively represent the redness and the bluishness of the color image. Therefore, the user must operate the apparatus by replacing the desired color and the extraction result with the features of reddish and bluish. That is,
The user judges the redness contained in the color to be extracted and R-
The work of setting the upper and lower limits of the window comparator for the Y input and observing the extraction results must be repeated for each input.

In addition, RY and B used at this time
Although the axes of −Y are substantially orthogonal to each other, the colors expressed by this coordinate system have a shape of a substantially polar coordinate system. That is, if the range is specified by the window comparator for each of the R-Y and B-Y input signals, the result is that the color is extracted including a color that the user does not intend.

Further, each input of RY and BY is
Since the brightness varies depending on the brightness of the input image, the conventional one has a drawback that the brightness of the input image is weak.

The present invention solves such a problem and enables the color setting and the identification extraction as intended by the user, and the color identification and extraction apparatus is less susceptible to the influence of the change in the brightness of the input image. The purpose is to provide.

[0007]

A color identification and extraction apparatus of the present invention identifies a color to be extracted manually by a color setting unit and a color set by this color setting unit from an input color video signal. In the color identification and extraction device including the identification and extraction means for extracting by the above, the identification and extraction means is a mapping means for mapping the input color video signal to a feature space having a color-related feature as each axis, and a mapping feature on the mapped feature space. A distance calculating means for obtaining a distance between the color set by the color setting means and the color represented by the input color video signal, and a detection output means for outputting a detection signal when the distance is equal to or less than a preset value. It is characterized by including and.

The input color video signal may be an RGB signal or any other signal such as RY, BY and Y.

As the preset value in the detection output means, the width for each axis in the feature space for the color set by the color setting means may be set, and the feature for the color set by the color setting means may be set. You may set the distance in space. When setting the distance in the feature space, the distance calculation means may include a means for calculating the distance in which the components of two or more axes are integrated in the feature space. Two
The Mahalanobis distance, for example, is used as the distance in which the components above the axis are integrated.

The color setting means includes means for designating the color to be set as a position on the screen represented by the input color video signal, and means for storing the output of the mapping means corresponding to the designated point. Is good.

It is desirable to further include a noise filter for removing a false color having a size of about one pixel by taking a majority decision of the output of the detection output means for each pixel and its surrounding pixels.

[0012]

The color identification and extraction device of the present invention maps a color video signal input from an image pickup device such as a color camera into a feature space having high orthogonality, and sets a range of colors to be extracted in the feature space. Identify and extract colors. As a result, the identification and extraction can be performed as intended by the user.
At this time, by selecting a feature space to be imaged whose brightness (luminance) is independent of other axes, stable color identification extraction that is not easily affected by changes in the brightness of the input image can be performed. . In this way, it is possible to improve the disadvantages of the conventional color identification / extraction device, and to realize a color identification / extraction device having far superior operability than the conventional type.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a block diagram showing a color identification extraction apparatus according to the first embodiment of the present invention.

The apparatus of this embodiment is provided with ROM tables 5, 6 and 7 as mapping means for mapping an input color video signal into a feature space having color-related features as individual axes, and the above-mentioned color on the mapped feature space. Distance calculation ROM tables 11 and 1 as distance calculation means for obtaining the distance between the color set by the setting means and the color represented by the input color video signal.
2 and 13, and magnitude comparators 14, 15 and 16 and an AND gate 17 are provided as detection output means for outputting a detection signal when the distance is equal to or less than a preset value.

Further, since the color to be extracted is manually set, the screen position designation circuit 19 is provided as a means for designating the color to be set as the position on the screen represented by the input color video signal. Data latches 8, 9 as means for storing the output of the mapping means corresponding to the
10 is provided.

The configuration of the apparatus of this embodiment will be described in more detail.

In this color discrimination extraction device, an R signal 21, a G signal 22 and a B signal 23 as input signals and a sync signal 3 are input.
1 and are input.

R signal 21, G signal 22 and B signal 23
Are converted into digital data by the A / D converters 1, 2 and 3, respectively, and the R terminal and G of the matrix calculator 4 are converted into digital data.
Input to the terminal and B terminal. Here, it is assumed that these digital data have an 8-bit structure. The matrix calculator 4 converts the input RGB data into R-Y, B-Y, and Y data and outputs it. The converted data also has an 8-bit structure. The RY and BY data output from the matrix calculator 4 are respectively RO
M tables 5 and 6 are entered. ROM table 5
Converts the input RY and BY data into visually-corrected hue data 24 and outputs it. The ROM table 6 obtains the saturation data before the correction by the brightness,
Output to ROM table 7. ROM table 7 at the same time
Is input with Y data from the matrix calculator 4. The ROM table 7 uses these data,
The saturation-corrected saturation data 25 is output. The Y data output by the matrix calculator 4 is the same as the lightness data 26.

The hue data 24, the saturation data 25, and the lightness data 26 are the distance calculation ROM table 1 respectively.
It is supplied to one input of 1, 12, 13 and further supplied to the data input of the data latches 8, 9, 10. The data outputs of the data latches 8, 9, 10 are connected to the other inputs of the distance calculation ROM tables 11, 12, 13, respectively. The outputs of the distance calculation ROM tables 11, 12, and 13 are the hue distance signal 27, the saturation distance signal 28, and the saturation distance signal 28, respectively.
The lightness distance signal 29 is supplied to one of the input terminals of the magnitude comparators 14, 15 and 16.

The sensitivity value setting circuit 20 includes a sensitivity setting volume and an A / D converter for converting the rotation angle thereof into digital data, and outputs a sensitivity value signal 33. The sensitivity value signal 33 is the magnitude comparator 14, 1
5 and 16 are supplied to the other input. magnitude·
The outputs of the comparators 14, 15 and 16 are AND gates 1.
7 inputs. The output of the AND gate 17 is the detection signal 30.

The sync signal 31 is input to the sync separation circuit 18. The output of the sync separation circuit 18 is the screen position designation circuit 19
Is supplied to the input of. The screen position designation circuit 19 outputs the designated color position pulse signal 32, and the data latches 8, 9, 10
Of each clock input.

Next, the operation of the apparatus of this embodiment will be described.

Input RGB signals 21, 22, 23
Is converted into digital data and input to the matrix calculator 4. The matrix calculator 4 is a digital circuit that performs a matrix calculation of 3 × 3, and Y = 0.299R + 0.587G + 0.114B (1) RY = 0.701R-0.587G-0.114B (2) BY = -0.299R-0.587G + 0.886B (3) is calculated. Actually, since the data that can be taken out as an output is composed of 8 bits, it is more advantageous in terms of accuracy when used in the subsequent stage that the output also has an amplitude of 0 to 255 when each RGB input changes from 0 to 255. . Therefore, scaling (normalization) for this is also performed at the same time, and Y = 0.299R + 0.587G + 0.114B (4) RY = 0.500R- 0.419G- 0.081B (5) BY = -0.169R- 0.331G + 0.500B (6) is calculated. At this stage, the brightness which is the first characteristic axis is separated from the input RGB data.

The remaining two characteristic axes are hue and saturation. Since these values are represented by a total of 16 bits, they can be used as inputs to the ROM table using a general-purpose memory. R
Although virtually any function can be realized by using the OM table, there is a drawback in that the number of input bits is limited. Therefore, in the present embodiment, first, the matrix calculator 4 separates one characteristic axis, and then the ROM table is used.

The hue is obtained from the ROM table 5. The ROM table 5 receives the RY and BY output from the matrix calculator 4, and outputs the visually corrected hue data 24. Specifically, H = tan ^-1 (BY) / (RY) (7) is used to calculate the hue H so that it matches the Munsell color system, that is, the visual sense, It is corrected as H '= f _H (H) (8).

The visual correction performed here will be described.
When the hue H calculated by the equation (7) is expressed by a color image, R (red), G (green) and B (blue) are 1 respectively.
It is placed at an angle of 20 degrees. However, the colors included in the 120 degrees are small between G and B, and considerably large between B and R and between R and G. This is consistent with the Munsell color system established by the human sense of color. Therefore, a function called f _H is set, and H is corrected so that the color density of the converted H ′ becomes constant.

FIG. 2 is a diagram for explaining the hue correction.
(A) shows a normal hue H, and (b) shows a corrected hue H '. In the normal hue H, the color density differs depending on the value (angle). This is corrected so that the color density becomes constant regardless of the value (angle) of the hue H '. H'corresponds to the Munsell color system.

Regarding the saturation, ROM tables 6 and 7
Ask by. The ROM table 6 inputs RY and BY output from the matrix calculator 4, and outputs the calculation result to the ROM table 7. ROM table 7
Receives the output of the ROM table 6 and the Y output from the matrix calculator 4, and outputs the brightness-corrected saturation S '. When the operation of the ROM table 6, and 7 represented by the formula, S = sqrt ((R- Y) 2 + (B-Y) 2) ... (9) S '= S × (1 / S max) × (Y max / Y) (10) However, sqrt is a function for obtaining a square root. In addition, the maximum value of the saturation for a certain hue H is S _max ,
The brightness when the saturation S is 1 is Y _max .

In the equation (10), except for the value of Y, RY and BY which are inputs to the ROM table 6 are used.
Can be calculated from Therefore, the ROM table 6, to calculate the _{S ROM1 = (1 / S max} ) × Y max × sqrt ((R-Y) 2 + (B-Y) 2) ... (11). The ROM table 7 calculates S _ROM2 = S _ROM1 / Y (12) using the calculation result of the expression (11) output from the ROM table 6. In this way, the two ROM tables 6
And 7, the brightness-corrected saturation is calculated.

The reason why the saturation is corrected by the brightness will be described. The saturation S that is normally used is an amount that depends on the brightness. For example, consider a pure color of R (red). At this time, B and G are 0, so by substituting the equation (4) into the equation (9), S = sqrt ((0.701R) ² + (0.299R) ² ) = 0.762R (13) . That is, when the value of R changes (the brightness changes), the saturation S also changes in proportion to this. Although the case of pure color is shown here, the same applies to the case of intermediate color. The brightness dependence of the saturation S is corrected by using the ROM table 7.

As described above, the input RGB is represented by the three axes (intensity-corrected hue, brightness-corrected saturation, and brightness) that are highly independent and have high linearity in accordance with human senses. Can be converted to the feature space. In this feature space, for example, with respect to color by human sense, the sense of difference in human sense directly corresponds to the numerical value of the difference of H ′. This is almost the same value for each color. Further, even if there is a change in the brightness when the aperture value of a color camera or the like is changed, the brightness-corrected saturation S ′ hardly changes, so that extremely stable color identification extraction can be performed.

Next, a method of designating a color in the color screen will be described. Conventionally, in order to set the range of hue and saturation while looking at the color extraction result, each knob is set to 2
I had to adjust four knobs, one by one. On the other hand, in this embodiment, the designated color is obtained from the color information captured at the designated position on the screen. In particular,
At the timing corresponding to the position on the screen specified by the user,
The color data after being converted into the feature space, that is, the hue data 24, the saturation data 25 and the lightness data 26 are latched in the data latches 8, 9 and 10, respectively. That is, the color information at the position on the screen designated by the user is stored in the data latches 8, 9 and 10.

For this purpose, the synchronizing signal 31 of the video signal is used. First, the input sync signal 31 is converted into a digital signal by the sync separation circuit 18. The synchronization signal includes, as information, a time difference from the left of the screen to an arbitrary point and a time difference from the screen to an arbitrary point. The screen position specifying circuit 19 specifies an arbitrary position on the screen by specifying these time differences. That is, the in-screen position is expressed as a time difference from the horizontal synchronizing signal and the vertical synchronizing signal by the pulse output from the screen position specifying circuit 19. With this pulse, the data latches 8, 9 and 10 latch each data input, and the visually corrected hue data 2
4. Brightness-corrected saturation data 25 and brightness data 2
Store 6 These color data are obtained by converting the color designated by the user into the feature space.

Next, the specified color and all Rs in the screen are displayed.
An operation for identifying and extracting only a close color within a certain distance from the color designated on the screen using the GB information will be described. In this embodiment, all the RGB information in the image is converted into the feature space in time series. That is, the RGB data of the pixels are sequentially converted into the information of the feature space according to the scanning of the video signal. The storage of the information transformed into the feature space has already been described. By calculating the distance in the feature space between the stored position in the feature space and the input image data that is sequentially converted in time series, the distance from the stored color, that is, the difference in color can be known. Furthermore, since the distance in the feature space at this time is one that is highly linear and orthogonal to the human sense as each axis of the feature space, the distance to the human sense of color, that is, the color It corresponds very well to the differences.

FIG. 3 is a diagram showing the relationship between the designated color and the extraction region. FIG. 3A shows the designated color in the feature space, and FIG.

Distance calculation ROM table 1 is used for distance calculation.
Use 1, 12, and 13. This is mainly to deal with the distance calculation regarding the hue. Since the hue data 24 is 8-bit digital data, it is represented by a numerical value from 0 to 255. However, since hue is an angle value, 0 is different from 255 by only 1. Since it is inconvenient to perform such a polar coordinate system operation only by a logic circuit, a ROM table is used in this embodiment. Distance calculation ROM
The table 11 obtains a distance related to hue and outputs a hue distance signal 27, the distance calculation ROM table 12 obtains a distance related to saturation and outputs a saturation distance signal 28, and the distance calculation ROM table 13 obtains a distance related to lightness. And outputs a lightness distance signal 29. These outputs are also 8-bit digital data.

The sensitivity of color identification extraction is set by the user operating the sensitivity value setting circuit 20. The selection sensitivity that the user wants to indicate is originally one. It is not familiar to human senses to specify a plurality of features such as a first sensitivity for hue and a second sensitivity for saturation. Therefore, in this embodiment, the set value relating to the sensitivity is set to one, and this is made to correspond to the operation member such as the volume. The sensitivity value setting circuit 20 outputs a sensitivity value signal 33 according to the user's setting. This is 8-bit digital data and is supplied to the magnitude comparators 14, 15 and 16.

Magnitude comparator 14, 1
Reference numerals 5 and 16 respectively compare the hue distance signal 27, the saturation distance signal 28, and the lightness distance signal 29 with the sensitivity value signal 33, and output signals when the distance is smaller than the set value. This output is supplied to the AND gate 17, and when the distance is smaller than the set value for all axes in the feature space,
The AND gate 17 outputs the detection signal 30. This detection signal 30 is also a time-series signal corresponding to the input video signal, and the detection signal 3 after the delay time of the circuit element from the time when the R signal 21, the G signal 22 and the B signal 23 are input.
0 is output.

Next, a method for associating the operation amount of the operation member such as the volume in the sensitivity value setting circuit 20 with the corresponding actual distance in the feature space will be described.

To the human sense, the change in hue is most noticeable, and the change in saturation is not so noticeable. That is, if the distance in the feature space for each of the hue and the saturation is used as it is, the influence of the distance on the saturation becomes large. For example, when the extraction sensitivity is set high (only the same color is extracted) by the sensitivity setting, the color may not be identified and extracted due to a subtle difference in saturation that cannot be distinguished by human senses. In order to correct such a difference between the human sense and the weight of each axis forming the feature space, the distance calculation ROM tables 11, 12, 1
3 multiplies the calculation result by a coefficient and outputs the result. For example, the distance calculation ROM table 12 multiplies the absolute value of the result of simple subtraction by a value smaller than 1 and outputs the distance in the feature space regarding saturation.

Further, by changing the coefficient value to be multiplied by the calculation result, it is possible to freely set the change amount of the distance in the feature space with respect to the rotation angle of the sensitivity setting volume. For example, when the sensitivity setting volume is set near the low sensitivity, the amount of change in the set distance in the feature space cannot be delicately set for the rotation angle. In the vicinity of such a set value, by increasing the coefficient value by which the calculation result is multiplied, the set value can be largely changed with a slight angle change of the sensitivity setting volume. On the contrary, near the high sensitivity setting, the distance setting for each of hue and saturation may be changed little by little so that the distance setting slightly changes with respect to the rotation angle of the volume. Such an example is shown in FIG. In this figure, (a) shows an example of an input image, (b) shows an extraction region when the identification extraction sensitivity is set low, and (c) shows an example of an extraction region when the identification extraction sensitivity is set high.

Regarding the lightness, the distance from the specified lightness is ignored except for the extremely low lightness and the extremely high lightness of the signals obtained by converting the input RGB signals.

FIG. 5 is a block diagram showing an example of a noise filter for removing noise included in the detection signal.

The input RGB signals, that is, the R signal 21, the G signal 22 and the B signal 23, in addition to the noise relating to the color contained in the background image, the noise (caused by the color filter used in front of the CCD of the color camera) ( False color) is also included. This is particularly remarkable when using a color camera that obtains a color image using one CCD. The reason is that the color filter is used on the front surface of the CCD, so that strictly speaking, the color information of the same point in the captured image cannot be obtained. For example, even if a landscape without a color is photographed, a fine color that does not actually exist may be observed at a position where the brightness (luminance) changes. Such a false color is also a mere color for the color identification and extraction device, and depending on the designated color, a detection signal is also output from this portion. Such false color occurs in a size of about one pixel.

Therefore, the detection signal 30 of the circuit shown in FIG.
Is input to the noise filter, and only the fine extraction noise is removed. The detection signal 30 is supplied to the data input of the D flip-flop 51 and the input of the 5-input majority logic circuit 55. The data output of the D flip-flop 51 is connected to the data input of the D flip-flop 52 and the input of the 5-input majority logic circuit 55. The data output of the D flip-flop 52 is connected to the data input of the D flip-flop 53 and the input of the 5-input majority logic circuit 55. The data output of the D flip-flop 53 is connected to the data input of the D flip-flop 54 and the input of the 5-input majority logic circuit 55. The clock signal 56 is supplied to the clock inputs of the D flip-flops 51 to 54.

That is, the D flip-flops 51 to 54 form a 4-bit shift register, and the detection signal 30
Are shifted at the timing of the clock signal 56. The input (detection signal 30) and output of this 4-bit shift register are input to the 5-input majority logic circuit 55. The 5-input majority decision logic circuit 55 makes a majority decision for 5 pixels in the time axis direction, and selects the logic occupying the majority of the inputs as an output. As a result, a small number of extracted noises are absorbed, and the noise-removed detection signal 57 is output.

FIG. 6 shows another configuration example of the noise filter.

The detection signal 30 of the circuit shown in FIG. 1 is supplied to the data input of the D flip-flop 61, the input of the 9-input majority logic circuit 69, and the input of the line delay element 63. The data output of the D flip-flop 61 is connected to the data input of the D flip-flop 62,
The flip-flops 61 and 62 form a shift register. The data outputs of the two D flip-flops 61 and 62 are connected to the inputs of the 9-input majority logic circuit 69.

The output of the line delay element 63 is the data input of the D flip-flop 64, and the 9-input majority logic circuit 69.
And the input of the line delay element 66. The data output of the D flip-flop 64 is connected to the data input of the D flip-flop 65, and these D flip-flops 64 and 65 form a shift register.
The data outputs of the two D flip-flops 64 and 65 are connected to the inputs of the 9-input majority logic circuit 69.

The output of the line delay element 66 is the data input of the D flip-flop 67 and the 9-input majority logic circuit 6.
9 inputs. The data output of the D flip-flop 67 is connected to the data input of the D flip-flop 68, and these D flip-flops 67 and 68 form a shift register. These two D flip-flops 6
The outputs of 7 and 68 are 9-input majority logic circuits 69.
Connected to the input of.

D flip-flops 61, 62, 64, 6
A clock signal 610 is provided to each of the clock inputs of 5, 67 and 68.

The D flip-flops 61 and 62 store the extraction signals for two pixels in the horizontal direction of the screen. Therefore, the detection signal 30 and the respective data outputs of the D flip-flops 61 and 62 can simultaneously process the extraction signals of three pixels in the horizontal direction of the screen. D flip-flops 64 and 65, D flip-flops 67 and 6
The same applies to 8.

The line delay elements 63 and 66 are each composed of a delay line or a memory element, and delay one line in the vertical direction of the screen. Therefore, the output of the line delay element 63 is located one line above the current pixel position. Further, the output of the line delay element 66 is at a position two lines above the current pixel position. Therefore, it is possible to process the extracted signals of three pixels on each line,
As a whole, a 3 × 3 pixel area can be processed, and the extracted information is simultaneously input to the 9-input majority logic circuit 69. The 9-input majority logic circuit 69 outputs the noise-removed detection signal 611 when the extraction signal is included in 5 or more pixels of 9 pixels that are simultaneously input, and does not output the detection signal 611 in the case of 4 pixels or less. Therefore, extremely effective noise removal can be performed.

FIG. 7 is a block diagram showing a color identification extraction apparatus according to the second embodiment of the present invention. This embodiment differs from the first embodiment in that the color identification extraction is performed using the distance in which the components of two or more axes in the feature space are integrated. Here, an example using the Mahalanobis distance will be described. The Mahalanobis distance is an extension of the ordinary concept of distance in Euclidean space to the concept in feature space, and is a quantity that expresses the proximity of features of an input signal. Only parts different from the first embodiment will be described below.

Hue distance signal 27 and saturation distance signal 28
Is input to the Mahalanobis distance calculation ROM table 71. The Mahalanobis distance calculation ROM table 71 performs Mahalanobis distance calculation in the feature space and outputs a Mahalanobis distance signal 73. The Mahalanobis distance signal 73 is supplied to one input of the magnitude comparator 72. The other input of the magnitude comparator 72 receives the sensitivity value signal 3 from the sensitivity value setting circuit 20.
3 is supplied. The output of the magnitude comparator 72 is connected to one input of the AND gate 17.

The brightness processing is the same as in the first embodiment, and the brightness distance signal 29 is supplied to one input of the magnitude comparator 16 and the other input of the magnitude comparator 16 is supplied from the sensitivity value setting circuit 20. Sensitivity value signal 33 is supplied. The output of the magnitude comparator 16 is connected to the other input of the AND gate 17.

FIG. 8 is a diagram for explaining the operation of this embodiment. FIG. 8A shows an example of a designated color in the feature space, and FIG. 8B shows an example of a range below the set distance centered on the designated color.

Mahalanobis distance calculation ROM table 71
Calculates the Mahalanobis distance in the feature space from the hue distance signal 27 and the saturation distance signal 28. Specifically, each of the hue distance signal 27 and the saturation distance signal 28 is multiplied by a constant and then squared to obtain the square root after addition. The constant to be multiplied at this time corresponds to the weight of the axis of each feature space. In this way, the Mahalanobis distance for hue and saturation is obtained.

In the case of this embodiment as well, as in the first embodiment, the sensitivity value setting circuit 20 has only one set value relating to sensitivity, and this corresponds to an operating member such as a volume. This set value is output as an 8-bit sensitivity value signal 33, and the magnitude comparators 72 and 16
Is supplied to. Magnitude comparator 72, 1
6 compares the Mahalanobis distance and the distance related to the brightness with a set value, and outputs a signal when the distance is smaller than the set value. This signal is input to the AND gate 17, and when the distance in the feature space is smaller than the set value,
The AND gate 17 outputs the detection signal 30.

In this embodiment, only the hue and the saturation are used for calculating the Mahalanobis distance in the feature space, but the lightness can also be used. For that purpose, a ROM table may be added. Furthermore, the constants for multiplying the hue and the saturation may be changed according to the lightness. For example, when the lightness is sufficiently high, both the hue and saturation feature values are adopted, and when the lightness is low (dark), the saturation is no longer reliable enough, and the hue is weighted. It can be performed.

In the above description, the influence of noise can be reduced by using the average value of the peripheral values of one point instead of sampling only one point as the designated color position. For example, it is possible to reduce the influence of noise included in the input image by averaging nine surrounding pixels including one designated point.

In the above embodiments, the structure for identifying and extracting only one color has been described. However, if this device is connected in parallel, a plurality of colors can be identified and extracted at the same time.

An example of using an 8-bit output A / D converter for digitizing an RGB signal has been described. However, by using only an A / D converter with a small number of bits or only the upper bits of the output signal, a matrix calculator is provided. It is also possible to replace the ROM table with the ROM table. In the current technology, the maximum ROM capacity is about 4 megabits, so the number of input signals at this time is 19. Therefore, for example, by using 6 bits for the R signal, 7 bits for the G signal, and 6 bits for the B signal, these are set to R
Used as the OM input signal. In this way, the circuit configuration becomes very simple. It is considered that the capacity of the ROM will be increased in the future, and practically sufficient gradation resolution can be obtained even with the ROM table type circuit configuration.

Although the example in which the distance calculation is performed by the ROM table has been described, the present invention can be similarly implemented by using other logic circuits. As for the saturation and brightness, since the numerical values are not cyclic, prepare two sets of normal subtraction circuits
And B-A may be calculated and the one with a positive sign may be adopted. The absolute value may be obtained by once converting into an analog signal and performing an analog operation.

As the noise filter, in addition to the examples described above, a noise filter in the time axis direction, an ordinary FIR digital filter, a combination thereof, and the like can be used.

Although the case where the input signal is an analog RGB signal has been described, the present invention can be similarly applied to the case of a signal of another format. For example, if the input signal is RY, B-
In the case of Y and Y signals, the matrix calculator is unnecessary.
Furthermore, it is also possible to directly input a digital signal as an input signal.

It is also extremely effective to perform preprocessing such as color temperature correction of the color signal before inputting the signal to the color identification and extraction device of the present invention. Also, the amplitude of the color signal can be automatically adjusted so that the color required by the user can be most effectively reproduced. For this purpose, it is preferable to adjust a lens diaphragm installed in a color camera or the like.

[0068]

As described above, the color extraction / identification device of the present invention converts an input color video signal into a feature space having high independence and linearity with respect to human senses, and the color is converted in the feature space. Identify and extract. As a result, it is possible to perform color identification extraction of an operation sensation that is stable with respect to changes in the brightness of the target and that also matches the human sensation well.

[Brief description of drawings]

FIG. 1 is a block configuration diagram showing a color identification extraction device of a first embodiment of the present invention.

FIGS. 2A and 2B are diagrams illustrating a hue correction, FIG. 2A is a normal hue H, and FIG. 2B is a diagram showing a corrected hue H ′.

FIG. 3 is a diagram showing a relationship between a designated color and an extraction area,
(A) is a figure which shows the designated color in a feature space, (b) shows the range below the set distance centering on it.

FIG. 4 is a diagram showing an operation example of color identification extraction, where (a) is an input image, (b) is an extraction region when identification extraction sensitivity is set low, and (c) is identification identification sensitivity set high. The figure which shows the example of the extraction area | region when it was performed.

FIG. 5 is a block diagram showing an example of a noise filter.

FIG. 6 is a block diagram showing another example of a noise filter.

FIG. 7 is a block configuration diagram showing a color identification extraction device of a second embodiment of the present invention.

FIG. 8 is a diagram for explaining the operation of the second embodiment, (a)
Is a designated color in the feature space, and (b) is a diagram showing a range below a set distance centered on the designated color.

[Explanation of symbols]

1, 2, 3 A / D converter 4 Matrix calculator 5, 6, 7 ROM table 8, 9, 10 Data latch 11, 12, 13 Distance calculation ROM table 14, 15, 16 Magnitude comparator 17 AND gate 18 Synchronization Separation circuit 19 Screen position designation circuit 20 Sensitivity value setting circuit 21 R signal 22 G signal 23 B signal 24 Hue data 25 Saturation data 26 Lightness data 27 Hue distance signal 28 Saturation distance signal 29 Lightness distance signal 30 Detection signal 31 Sync signal 32 designated color position pulse signal 33 sensitivity value signal 51 to 54 D flip-flop 55 5 input majority logic circuit 56 clock signal 57 noise-removed detection signal 61, 62, 64, 65, 67, 68 D flip-flop 63, 66 lines Delay element 69 9-input majority logic circuit 610 Lock signal 611 Noise-free detection signal 71 Mahalanobis distance calculation ROM table 72 Magnitude comparator 73 Mahalanobis distance signal

Claims (7)

[Claims] 1. Color identification extraction comprising color setting means for manually setting a color to be extracted and identification and extraction means for identifying and extracting the color set by the color setting means from an input color video signal. In the apparatus, the identification and extraction means includes a mapping means (5, 6, 7) for mapping an input color video signal onto a feature space having color-related features as individual axes, and the color setting means on the mapped feature space. Distance calculating means (11, 12, 13) for obtaining the distance between the color set by the above and the color represented by the input color video signal, and a detection signal is output when this distance is less than or equal to a preset value. A color identification and extraction device comprising a detection output means (14, 15, 16, 17). 2. The color identification and extraction device according to claim 1, wherein the characteristics relating to the color are hue, saturation and lightness. 3. A means (20) for setting a width for each axis in the feature space for the color set by the color setting means as the preset value in the detection output means. Item 1. The color identification extraction device according to item 1 or 2. 4. A means (20) for setting a distance in the feature space for a color set by the color setting means as the preset value in the detection output means, the distance calculation means The color identification extraction device according to claim 1 or 2, further comprising: means (72) for calculating a distance in which the components of two or more axes in the feature space are integrated. 5. The color identification extraction device according to claim 4, wherein the distance in which the components of two or more axes are integrated is a Mahalanobis distance. 6. The color setting means specifies the color to be set as a position on the screen represented by the input color video signal, and the output of the mapping means corresponding to the specified point. 6. The color identification and extraction device according to claim 1, further comprising means (8, 9, 10) for storing. 7. A noise filter for removing a false color having a size of about one pixel by taking a majority decision of the output of the detection output means for each pixel and its surrounding pixels. Any one of the color identification extraction devices.