JP5598185B2 - Conspicuous image generating apparatus and conspicuous image generating program - Google Patents

Description

  The present invention relates to a technique for evaluating the degree of conspicuousness of an object.

Conventionally, as a technique for evaluating the light environment created by illumination or natural light, a brightness distribution image showing the brightness distribution of the light environment is generated based on the captured image of the light environment to be evaluated, and this brightness distribution image is referenced. A technique for evaluating the light environment is known.
In recent years, there has also been known a technique for generating a brightness image that quantitatively shows the distribution of brightness perception felt by humans from the brightness distribution image and evaluating the light environment based on the brightness perception image. Yes. In this technology, based on the correspondence between predetermined brightness and brightness perception, the brightness of the brightness distribution image is converted into brightness perception to generate an image that quantitatively represents the brightness perception of the evaluation target. (For example, refer to Patent Document 1).

International Publication No. 2006/132014 Pamphlet

However, whether or not the evaluation object is easily perceived by humans in the light environment, that is, whether or not the evaluation object is conspicuous in the light environment cannot be determined only by the perception of the brightness of the evaluation object.
For example, even when the luminance of the evaluation target is low, when the color of the evaluation target is different from the surroundings, it can be perceived that the evaluation target is conspicuous depending on the difference in the color.
More specifically, when a surface of the same color having uniform high brightness extends over the entire field of view, brightness can be perceived, but noticeable perception is not detected. This is because, in conspicuousness, the absolute amount of luminance is not important, and the presence or absence of comparison is important.
The present invention has been made in view of the above-described circumstances, and an object thereof is to provide a conspicuous image generation device and a conspicuous image generation program that can quantitatively evaluate the conspicuousness of an evaluation object by reflecting a difference in color. .

In order to achieve the above object, the present invention provides a uniform color designed so that the Euclidean distance between two points between coordinates corresponds to a perceptual color difference between the values of each pixel of a captured image obtained by capturing an evaluation target. degree view, or converted into the coordinates of a uniform color space, and evaluated that there is no noticeable when the color difference based on the Euclidean distance of the evaluated coordinates and surrounding coordinate is not, the coordinates coordinates and the periphery of the evaluation When the color difference contrast amount based on the Euclidean distance is large, the color difference conspicuous evaluation unit that highly evaluates the conspicuousness of the evaluation object according to the degree of the size, and the imaging based on the evaluation result of the color difference conspicuous evaluation unit A conspicuous image generating apparatus comprising: an imaging unit that generates a conspicuous image indicating a conspicuous distribution due to color difference of an image.

The present invention, in the above-mentioned noticeable image generating apparatus, the color difference noticeable evaluation unit, the color difference of the evaluation object and the surroundings, even the chromaticity diagram by Wavelet analysis, or comparison of each coordinate component of the coordinate of the uniform color space after calculating seeking amount as the Euclidean distance between the coordinates, determine the Euclidean distance contrast amount representing the chrominance contrast amount by Wavelet analysis, and evaluating the conspicuous based on the Euclidean distance contrast amount.

  According to the present invention, in the conspicuous image generation apparatus, the captured image obtained by capturing the evaluation object is evaluated as being inconspicuous when the evaluation object and the surrounding luminance are equal, and the evaluation object is compared with the surrounding luminance. A luminance conspicuous evaluation unit that highly evaluates the conspicuousness of the evaluation object according to the degree of the size when the amount is large, and the imaging unit evaluates the color difference conspicuous evaluation unit and the luminance conspicuous evaluation unit Based on the result, a conspicuous image showing a distribution of conspicuous color difference and luminance of the captured image is generated.

In order to achieve the above object, the present invention is designed such that the value of each pixel of a captured image obtained by capturing an evaluation target is set so that the Euclidean distance between two points between coordinates corresponds to a perceptual color difference. Converted to a uniform chromaticity diagram or coordinates of a uniform color space, and evaluated that there is no noticeable color difference based on the Euclidean distance between the coordinates of the evaluation target and the surrounding coordinates, and the coordinates of the evaluation target and the coordinates When the color difference contrast amount based on the Euclidean distance of the surrounding coordinates is large, based on the evaluation result of the color difference conspicuous evaluation unit, means for highly evaluating the conspicuousness of the evaluation object according to the degree of the size, A conspicuous image generation program is provided that functions as means for generating a conspicuous image indicating a conspicuous distribution due to a color difference of the captured image.

  According to the present invention, for a captured image obtained by imaging the evaluation object, it is evaluated that there is no noticeable color difference between the evaluation object and the surrounding area, and when the evaluation object and the surrounding color difference contrast amount are large, In accordance with the degree of the size, the evaluation object is highly evaluated for its conspicuousness, and based on the evaluation result, a conspicuous image reflecting the conspicuousness is generated from the captured image. It is possible to quantitatively grasp how conspicuous from the conspicuous image by the difference.

It is a figure which shows the structure of the conspicuous evaluation system which concerns on embodiment of this invention. It is a block diagram which shows the functional structure of a conspicuous image generation apparatus. It is a block diagram which shows the functional structure of a color difference conspicuous evaluation part. It is a figure which shows the relationship between a pixel and a surrounding pixel. It is explanatory drawing of (DELTA) u 'intermediate image, (DELTA) v' intermediate image, and (DELTA) u'v 'color difference image. It is a block diagram which shows the functional structure of a brightness conspicuous evaluation part. It is a block diagram which shows the functional structure of an imaging part. It is a figure which shows the sample of the stimulus used for the magnitude estimation method. It is a figure which shows an example of the evaluation result by a magnitude estimation method when the size and chromaticity of a target area are changed, (A) changes the size of a target area, (B) shows the case where chromaticity is changed. Show. It is a figure which shows an example of the evaluation result by a magnitude estimation method when the size and luminance ratio of a target area are varied, (A) is a case where the size of a target area is varied, (B) is a case where a luminance ratio is varied. Show. It is a coordinate and brightness | luminance of the xy chromaticity diagram of the color used for the test. It is a flowchart of a conspicuous image generation process. It is a figure for demonstrating wavelet decomposition. It is a figure which shows an example of contrast with the surrounding pixel using wavelet decomposition. It is a figure for demonstrating the production | generation of the conspicuous image according to the level of wavelet decomposition. It is a sample figure imitating the picked-up image data which imaged the light environment which arranged the disk on the background of one color, and shows the luminance distribution of each picked-up image data while showing the size and color of the disk in a variable manner. is there. It is a sample figure of a conspicuous image which evaluated conspicuousness about the picked-up image data in FIG. It is a flowchart of the conspicuous image generation process which concerns on the modification of this invention. It is a flowchart of the conspicuous image generation process which concerns on the other modification of this invention. It is a flowchart of the conspicuous image generation process which concerns on the other modification of this invention.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram illustrating a configuration of a conspicuous evaluation system 1 according to the present embodiment.
The conspicuous evaluation system 1 includes a camera 7, a conspicuous image generating device 11, and a display device 13.
The camera 7 captures the light environment 3 including the evaluation target of conspicuousness and inputs digital data of the captured image (hereinafter referred to as “captured image data 5”) to the conspicuous image generating device 11. As the captured image data 5, data obtained by reading a print image of the captured image of the light environment 3 with a scanner device and converting it into digital data can be used. Alternatively, the captured image data 5 may be data obtained by reproducing the light environment 3 with computer graphics (CG) and rendering it.

The conspicuous image generating device 11 generates conspicuous image data 9 based on the captured image data 5 and outputs it to the display device 13. The conspicuous image data 9 is conspicuous image data that quantifies and shows the conspicuousness of the evaluation target included in the light environment 3. As the conspicuous image, a color mapping image is used in which the degree of conspicuousness felt when a person visually recognizes the evaluation target is expressed by coloring each region using a pseudo color.
The display device 13 displays a conspicuous image based on the conspicuous image data 9, and this display allows the user to quantitatively grasp the conspicuousness of the evaluation target of the light environment 3. Note that the conspicuous image data 9 may be output to a printer device to be imaged.

FIG. 2 is a block diagram illustrating a functional configuration of the conspicuous image generation apparatus 11.
The conspicuous image generation apparatus 11 includes an image input unit 21, a conspicuous evaluation unit 23, an imaging unit 25, and an image output unit 27. The image input unit 21 is connected to the camera 7 that outputs the captured image data 5, and outputs the captured image data 5 output from the camera 7 to the conspicuous evaluation unit 23. In addition to the camera 7, any device that outputs the captured image data 5 can be connected to the image input unit 21. The conspicuous evaluation unit 23 evaluates the conspicuousness of the evaluation object based on the captured image data 5 and outputs the conspicuous evaluation result to the imaging unit 25. The imaging unit 25 generates the conspicuous image data 9 obtained by imaging the conspicuity of the evaluation target based on the conspicuous evaluation result, and outputs the conspicuous image data 9 to the image output unit 27. The image output unit 27 outputs the conspicuous image data 9 to the display device 13.
Further, the conspicuous evaluation unit 23 evaluates conspicuousness based on a color difference conspicuous evaluation unit 29 that evaluates conspicuousness based on the color difference between the evaluation target color and surrounding colors, and a contrast between the evaluation target luminance and the surrounding luminance. And a brightness conspicuousness evaluation unit 30 that performs conspicuousness evaluation comprehensively using both the color and brightness of the evaluation target.
The conspicuous image generation apparatus 11 is implemented by causing a so-called computer having a CPU, RAM, and ROM to execute a conspicuous image generation program 15 (FIG. 1) for realizing the above functional blocks. .

FIG. 3 is a diagram illustrating a functional configuration of the color difference conspicuous evaluation unit 29.
The color difference conspicuous evaluation unit 29 evaluates the conspicuousness by converting the value of each pixel of the captured image, which is a color image (RGB information), into color coordinate information in order to evaluate the conspicuousness based on the color difference between the evaluation target and the surroundings. To do. As the color coordinate information, a uniform chromaticity diagram or coordinate information of a uniform color space designed so that the distance between two points between coordinates corresponds to a perceptual color difference is used. In the present embodiment, coordinate information of a u′v ′ chromaticity diagram that is a uniform chromaticity diagram is used.

The functional configuration of the color difference conspicuous evaluation unit 29 will be described. As shown in FIG. 3, the color difference conspicuous evaluation unit 29 includes a u ′ image generation unit 31 and a v ′ image generation unit 33, and a first wavelet decomposition unit. 35, a Δu ′ intermediate image generation unit 36 and a Δv ′ intermediate image generation unit 37, a Δu′v ′ color difference image generation unit 38, a second Wavelet decomposition unit 39, and a coefficient processing unit 40.
The u ′ image generation unit 31 generates a u ′ image from the captured image data 5, and the v ′ image generation unit 33 generates a v ′ image from the captured image data 5 and outputs the v ′ image to the first Wavelet decomposition unit 35. The u ′ image generation unit 31 and the v ′ image generation unit 33 each of the captured image data 5 based on approximate matrix information for converting RGB information acquired from the camera 7 into chromaticity coordinates (u ′, v ′). The RGB value of each pixel is converted into an image decomposed into u ′ chromaticity component and v ′ chromaticity component. The approximate matrix information is obtained by a calibration experiment and is stored in advance in the color difference conspicuous evaluation unit 29.

  The first Wavelet decomposition unit 35, for each pixel of the u ′ image and the v ′ image, each of a pixel (hereinafter referred to as “surrounding pixel”) surrounding the pixel (hereinafter referred to as “target pixel”). The chromaticity component contrast amount (u ′ contrast amount and v ′ contrast amount) reflecting the chromaticity component difference is calculated. As shown in FIG. 4, the surrounding pixel P is defined as a pixel located on the circumference of a circle having a radius R centered on the target pixel O. That is, the chromaticity component contrast of the target pixel O and the chromaticity component composed of the surrounding pixels P surrounding the target pixel O are obtained as the chromaticity component contrast amount. At this time, in each of the u ′ chromaticity component and the v ′ chromaticity component, when the target pixel O has a larger value than the surrounding pixel P (positive contrast), the sign of the chromaticity component contrast amount becomes positive and the value is small. In this case (reverse contrast), the sign of the chromaticity component contrast amount is negative. Further, when the chromaticity component contrast amount of the target pixel O is large with respect to all of the surrounding pixels P, the absolute value of the chromaticity component contrast amount is maximum, and the chromaticity component contrast amount with the target pixel O is small. Is included in the surrounding pixels P, the absolute value becomes smaller. Further, when all the surrounding pixels P have the same value as the target pixel O, the absolute value of the chromaticity component contrast amount is minimized.

  Further, the value of the chromaticity component contrast amount (u ′ contrast amount and v ′ contrast amount) also varies depending on the distance between the target pixel O and the surrounding pixel P. For example, assuming a state in which pixels having the same value continue around the target pixel O, when a pixel within this continuous range is set as the surrounding pixel P, the absolute value of the chromaticity component contrast amount becomes small. On the contrary, when the surrounding pixel P is set outside the continuous range, the absolute value of the chromaticity component contrast amount increases. In order to reflect the difference in distance between the target pixel O and the surrounding pixel P in the chromaticity component contrast amount, in this embodiment, the radius R is set to levels 1 to N (N = 9 in this embodiment). It is defined in stages, and a chromaticity component contrast amount is obtained for each level. Note that a range in which the same pixel values are continuous is estimated as one object shown in the captured image. That is, it can be said that obtaining the chromaticity component contrast amount by dividing the radius R into levels 1 to N obtains the chromaticity component contrast amount according to the size of the evaluation target.

  The first Wavelet decomposition unit 35 calculates the amount of comparison for each level using Wavelet analysis, and generates N Wavelet decomposition images (detailed images) for each level for each u ′ image and v ′ image. In this detailed image, the value of each pixel corresponds to a chromaticity component contrast amount.

By the way, in the chromaticity coordinates, unlike the luminance, the conspicuous code division based on the difference between the direct contrast and the reverse contrast does not make sense. That is, in the luminance contrast, when the luminance of the target pixel O is larger than the luminance of the surrounding pixel P, the target pixel O is bright and conspicuous (contrast contrast). Conversely, the luminance of the target pixel O is the surrounding pixel. When the luminance is lower than P, the target pixel O is dark and conspicuous (reverse contrast). For this reason, in luminance contrast, it is meaningful to distinguish whether the conspicuity depends on the positive contrast or the reverse contrast by the sign of positive or negative. On the other hand, in each of the u ′ chromaticity component and the v ′ chromaticity component, the difference between the direct contrast and the reverse contrast merely represents the positional relationship on the chromaticity coordinates of the target pixel O and the surrounding pixel P. It is not a difference that directly affects the evaluation of conspicuousness. In other words, in the chromaticity coordinates, if the contrast amount of the target pixel O and the surrounding pixel P is left with a positive or negative sign, it becomes an evaluation of different conspicuousness even if it is the same conspicuous perception. It is necessary to evaluate the conspicuousness based on the absolute value of.
Therefore, the absolute value of the difference between the target pixel O and the surrounding pixel P in each of the u ′ image and the v ′ image is calculated as the chromaticity component contrast amount, and the detail image is based on the chromaticity contrast amount Q. The method of generating

  However, in the Wavelet analysis, if all the negative values in the detail image are converted to positive values, an image before Wavelet decomposition (the u ′ image and the v ′ in the present embodiment) is obtained even if Wavelet synthesis is performed thereafter. (Image) will not return. That is, if the color conspicuousness is evaluated based on the detail image converted to all positive values, it is difficult to say that the color conspicuousness in the captured image data 5 is evaluated, and an accurate evaluation result cannot be obtained.

Therefore, in the present embodiment, instead of evaluating color conspicuousness based on the chromaticity component contrast amounts (u ′ contrast amount and v ′ contrast amount) of the u ′ image and the v ′ image, the target pixel is not evaluated. The conspicuousness of the color is evaluated based on the Euclidean distance D on the u′v ′ chromaticity diagram indicating the color difference between O and the surrounding pixel P.
The Euclidean distance D is obtained by the following equation (1) using the difference Δu ′ in the u ′ image between the target pixel O and the surrounding pixel P and the difference Δv ′ in the v ′ image.

In Expression (1), subscripts i and j are variables indicating the pixel position coordinates (FIG. 5) in the u ′ image and the v ′ image, respectively.

  As is clear from Equation (1), the Euclidean distance D does not have a negative sign. For this reason, it is not necessary to convert the detail image including the negative value of the u ′ image and the v ′ image into a positive sign and evaluate the conspicuousness, and therefore the conspicuousness due to the color difference can be accurately evaluated.

The configuration for the conspicuous evaluation using the Euclidean distance D will be described. In FIG. 3, the Δu ′ intermediate image generation unit 36 of the color difference conspicuous evaluation unit 29 uses 1 of the u ′ image that has been wavelet transformed by the first Wavelet decomposition unit 35. The N-level detail image is wavelet recombined excluding the approximate image (bias), and output to the Δu′v ′ color difference image generation unit 38. Similarly, the Δv ′ intermediate image generation unit 37 re-synthesizes the Wavelet 1-N level detailed images of the v ′ image, excluding the approximate image (bias), and outputs the result to the Δu′v ′ color difference image generation unit 38. To do. Wavelet transform is a process of analyzing the amount of change in signal at each resolution (frequency), so-called multi-resolution analysis. Therefore, re-synthesis is performed by removing approximate images (bias) from 1 to N level detail images. By doing so, it becomes possible to extract only the change of the signal at each level. Specifically, as shown in FIG. 5A, wavelet recombination is performed except for the approximate image (bias) even if the values of the surrounding pixels P are different in each of the u ′ image and the v ′ image. Thus, as shown in FIG. 5B, the values of the surrounding pixels P are both set to a predetermined constant value (zero in this embodiment) to express the differences Δu ′ and Δv ′ between the target pixel O and the surrounding pixels P. Images can be generated as Δu ′ intermediate images and Δv ′ intermediate images.
As shown in FIG. 5C, the Δu′v ′ color difference image generation unit 38 is based on the Δu ′ intermediate image and the differences Δu ′ and Δv ′ between the pixels of the Δv ′ intermediate image and the equation (1). The second wavelet is generated in order to generate a Δu′v ′ color difference image 81 in which the value of the target pixel O is displayed at the above-described Euclidean distance D, and to evaluate the color conspicuity by wavelet analysis based on the Δu′v ′ color difference image 81. The data is output to the decomposition unit 39.

  Similar to the first Wavelet decomposition unit 35, the second Wavelet decomposition unit 39 generates Wavelet decomposition images of 1 to N levels by Wavelet analysis on the Δu′v ′ color difference image 81, and uses these N Wavelet decomposition images as coefficients. The data is output to the processing unit 40. Each pixel value of the Wavelet decomposition image at each level indicates the Euclidean distance contrast amount Q at that level. The Euclidean distance contrast amount Q indicates the contrast amount of the color difference between the target pixel O and the surrounding pixel P at that level. Further, as will be described in detail later, by calculating the Euclidean distance contrast amount Q using the Wavelet decomposition, in addition to the comparison of the Euclidean distance between the target pixel O and the surrounding pixel P, the target pixel O and the surrounding pixel P can be compared. A value reflecting the contrast of the Euclidean distance between each pixel existing in between and the target pixel O is obtained as the Euclidean distance contrast amount Q.

The coefficient processing unit 40 performs coefficient processing for converting the Euclidean distance contrast amount Q of each pixel of the N wavelet decomposed image data output from the second wavelet decomposition unit 39 into a value indicating the degree of conspicuousness, and performs wavelet composition, and color difference The image is output to the imaging unit 25 as a conspicuous image 85. As the coefficient processing, in the Wavelet decomposed image of each level of 1 to N, each of the Euclidean distance contrast amounts Q1 to QN of each pixel is multiplied by multiple regression coefficients α1 to αN corresponding to the level. As a result, image data obtained by converting the value of each pixel of the Wavelet decomposition image into the intermediate generation values IQ1 to IQN of the Euclidean distance contrast amount Q is generated for each level.
The intermediate generation values IQ1 to IQN are values indicating the degree of conspicuousness of the Euclidean distance contrast amount Q of each pixel, and are values obtained by the magnitude estimation method. Details of the multiple regression coefficient α and the magnitude estimation method will be described later.
Then, the coefficient processing unit 40 wavelet synthesizes the N wavelet decomposed image data subjected to the coefficient processing, and outputs the result to the imaging unit 25 as a color noticeable image 85. The color noticeable image 85 is an image obtained by converting the value of each pixel of the Δu′v ′ color difference image into a value obtained by adding the intermediate generation values IQ1 to IQN of each level (hereinafter referred to as “color difference noticeable predicted value MQ”). It corresponds to.

  In this way, a Δu′v ′ color difference image 81 in which the color difference between the target pixel O and the surrounding pixel P is represented by the Euclidean distance D on the u′v ′ chromaticity diagram is generated, and this Δu′v ′ color difference image 81 is generated. A wavelet analysis including recombination and wavelet decomposition was performed on the image to evaluate color conspicuousness. As a result, it is not necessary to align the signs (take absolute values) in order to eliminate the difference in the sign due to the difference between the direct contrast and the reverse contrast on the chromaticity coordinates of the target pixel O and the surrounding pixel P. Accurate evaluation is possible.

Further, when generating the color difference conspicuous image 85 by wavelet analysis for each of the u ′ image and the v ′ image, the u ′ chromaticity component contrast amount of the u ′ image and the v ′ chromaticity component contrast amount of the v ′ image. The multiple regression coefficient α is required for each of the above. For this reason, the color conspicuousness is evaluated using the multiple regression coefficient α, which is twice the number of decomposition levels, as an explanatory variable.
In contrast, by using the Euclidean distance D, the wavelet analysis target image can be only the Δu′v ′ color difference image 81, and the number of multiple regression coefficients α is equal to the number of decomposition levels. And the reliability per explanatory variable can be increased.

FIG. 6 is a block diagram illustrating a functional configuration of the luminance conspicuous evaluation unit 30.
The luminance conspicuous evaluation unit 30 includes a log luminance distribution image generation unit 45, a Wavelet decomposition unit 47, and a coefficient processing unit 49.
The logarithmic luminance distribution image generation unit 45 generates a luminance distribution image indicating the luminance distribution of the light environment 3 based on the captured image data 5, and takes the logarithm of the luminance value of each pixel of the luminance distribution image to obtain a logarithmic luminance distribution image. Is output to the Wavelet decomposition unit 47.

  For example, photographic photometry is used to generate a luminance distribution image so that luminance values can be generated in a wide luminance range. That is, the logarithmic luminance distribution image generation unit 45 acquires a plurality of captured image data 5 obtained by capturing the light environment 3 with different exposure conditions (for example, shutter speed), and for each captured image data 5, each pixel of the captured image. Luminance data is generated by converting the gradation value of the value into a luminance value, and the luminance data of each captured image data 5 is synthesized to generate a luminance distribution image.

  The Wavelet decomposition unit 47 uses the Wavelet decomposition in the same manner as the first Wavelet decomposition unit 35 of the color difference conspicuous evaluation unit 29 to reflect the luminance contrast of each pixel of the log luminance distribution image with each of the surrounding pixels P of the pixel. The luminance contrast amount L is calculated for each of a plurality of levels 1 to N (N = 9 in this embodiment) corresponding to the distance to the surrounding pixel P. The value of the brightness contrast amount L is positive when the brightness of the target pixel O is higher than that of the surrounding pixels P (positive contrast), and is negative when the brightness is low (reverse contrast). In addition, the luminance contrast amount L has a maximum absolute value when the luminance contrast of the target pixel O is large with respect to all the surrounding pixels P, and a pixel having a small luminance contrast with the target pixel O is the surrounding pixel P. The absolute value becomes smaller as it is included. Further, when all the surrounding pixels P have the same luminance as the target pixel O, the absolute value of the luminance contrast is minimized.

The coefficient processing unit 49 wavelet synthesizes the N pieces of image data output from the wavelet decomposition unit 47, and outputs the result to the imaging unit 25 as a luminance noticeable image 86.
The coefficient processing unit 49 performs coefficient processing for converting the luminance contrast amount L of each pixel of the N wavelet decomposed image data output from the wavelet decomposition unit 47 into a numerical value indicating the degree of conspicuousness, and wavelet synthesizes the luminance noticeable image. 86 is output to the imaging unit 25. As the coefficient processing, in the Wavelet decomposed image of each level of 1 to N, each of the luminance contrast amounts L1 to LN of each pixel is multiplied by multiple regression coefficients β1 to βN for the luminance contrast amount L corresponding to the level. . As a result, image data in which the value of each pixel of the Wavelet decomposition image is converted into intermediate generation values IL1 to ILN for luminance comparison is generated for each level.
The intermediate generation values IL1 to ILN are values indicating the degree of conspicuousness of the luminance contrast amount L of each pixel, and are values obtained by the magnitude estimation method. Details of the multiple regression coefficient β and the magnitude estimation method will be described later.
Then, the coefficient processing unit 49 wavelet synthesizes the N wavelet decomposed image data subjected to the coefficient processing, and outputs the result to the imaging unit 25 as a luminance noticeable image 86. The luminance conspicuous image 86 corresponds to an image obtained by converting the value of each pixel of the Wavelet decomposition image data into a value obtained by adding the intermediate generation values IL1 to ILN of each level (hereinafter referred to as “luminance conspicuous prediction value ML”). To do.

FIG. 7 is a block diagram showing a functional configuration of the imaging unit 25.
The imaging unit 25 combines the color conspicuous image 85 output from the color difference conspicuous evaluation unit 29 and the luminance conspicuous image 86 output from the luminance conspicuous evaluation unit 30 to obtain the color difference between the target pixel O and the surrounding pixels P and A conspicuous image indicating a conspicuous degree in which both luminance differences are determined in an integrated manner is generated and output to the image output unit 27. Specifically, the imaging unit 25 includes a luminance conspicuous prediction value code conversion unit 65, a color difference / luminance conspicuity synthesis unit 67, a multistage evaluation value conversion unit 69, and an evaluation value code conversion unit 71. .

The luminance conspicuous prediction value code conversion unit 65 takes the absolute value of the luminance conspicuous prediction value ML at each pixel of the luminance conspicuous image in order to enable the summation of the color difference conspicuous and the luminance conspicuous without taking a negative sign. The predicted luminance conspicuation value ML is changed to a positive sign and is output to the luminance / color difference conspicuous composition unit 67.
The color difference / luminance conspicuous composition unit 67 adds the luminance conspicuous predicted value ML of the luminance conspicuous image and the color difference conspicuous predicted value MQ of the color conspicuous image 85 for each pixel, and converts it into a conspicuous value U reflecting the color and luminance. And output to the multistage evaluation value conversion unit 69.

  In the present embodiment, the imaging unit 25 obtains the conspicuous value U by adding the color difference conspicuous predicted value MQ and the luminance conspicuous predicted value ML, but is not limited thereto. That is, any function f can be used as long as the function f sufficiently satisfies the conspicuous value U≈f (color difference conspicuous predicted value MQ, luminance conspicuous predicted value ML).

The multi-level evaluation value conversion unit 69 converts the conspicuous value U of each pixel of the conspicuous image into a multi-level evaluation value (13 levels in the present embodiment), and associates each evaluation value with a pseudo color to color the degree of conspicuity. A conspicuous image that is a color mapping image expressed as described above is generated and output to the evaluation value code converter 71.
The evaluation value code conversion unit 71 assigns a negative sign to the multi-stage evaluation value of the pixel that has been positively encoded by the luminance conspicuous prediction value code conversion unit 65, and generates a conspicuous image that distinguishes the normal contrast and the reverse contrast of the luminance. And output to the image output unit 27. Thereby, the conspicuous image data 9 of the conspicuous image indicating the conspicuous distribution of the captured image is output from the image output unit 27 to an external output device such as the display device 13.

  Here, for the conversion from the conspicuous value U to the multi-stage evaluation value, a relationship obtained in advance through experiments is used. More specifically, the evaluation stimulus is presented to the subject, and the conspicuous feeling felt from the evaluation stimulus is “very conspicuous”, “well conspicuous”, “conspicuous”, “slightly conspicuous”, “slightly conspicuous” 7 evaluation phrases of “inconspicuous” and “invisible”, and 13 grades in the middle of each of these evaluation phrases. This evaluation stimulus is a stimulus whose brightness noticeable prediction value ML and color difference noticeable prediction value MQ (that is, the noticeable value U) are known. Such an evaluation test is performed on a large number of subjects using a large number of evaluation stimuli, and the correlation between the conspicuous value U of these evaluation stimuli and the 13-level evaluation is obtained. This correlation is used for the conversion from the conspicuous value U to the 13-level evaluation value by the multi-level evaluation value conversion unit 69.

Next, the above-described magnitude estimation method employed in this embodiment in order to obtain the correlation between the Euclidean distance contrast amount Q and the color difference noticeable prediction value MQ and the correlation between the brightness contrast amount L and the brightness noticeable prediction value ML will be described. .
In the magnitude estimation method, first, the subject is allowed to observe the reference stimulus. The subject is instructed beforehand so that the degree of conspicuousness of the reference stimulus is 100 points. Next, an evaluation stimulus was presented to the subject, and the degree of conspicuousness of the evaluation stimulus was evaluated by a score. At this time, an evaluation of 0 points is assumed to respond only when the target of the evaluation stimulus cannot be visually recognized, and a negative value is not used for the evaluation. There was no particular upper limit for the evaluation score. The score obtained by the magnitude estimation method is a conspicuous predicted value for the evaluation stimulus.

As shown in FIG. 8 (A), the reference stimulus has a viewing angle at the center of a background area 70 of a square (17 [cd / m ² ]) with a viewing angle of 90.0 [deg] on one side and a diameter of 20. A sample in which a circular target region 72 of 0 [deg] (luminance 50 [cd / m ² ]) was drawn was used. The background area 70 and the target area 72 are both achromatic.
In the experiment for obtaining the correlation between the Euclidean distance contrast amount Q and the color difference conspicuous predicted value MQ, the influence of the difference in the color difference (Euclidean distance) between the background region 70 and the target region 72 and the size of the target region 72 is noticeable. Therefore, the chromaticity (for example, FIG. 8B) and the size of the target region 72 are presented to the subject as evaluation stimuli. At this time, in order to remove the influence of the luminance contrast, the luminance of both the reference stimulus and the evaluation stimulus was adjusted so that the background region 70 and the target region 72 were both 20.0 [cd / m ² ].
Further, in the experiment for obtaining the correlation between the luminance contrast amount L and the luminance conspicuous predicted value ML, in order to examine the influence of the luminance ratio between the background region 70 and the target region 72 and the difference in the size of the target region 72 on the target, What changed the brightness | luminance and size of the area | region 72 (for example, FIG.8 (C)) was shown to the test subject as an evaluation stimulus. At this time, in order to remove the influence of the color difference, the background region 70 and the target region 72 are both achromatic (color 7 in FIG. 11) in both the reference stimulus and the evaluation stimulus.

  The evaluation test by the magnitude estimation method is performed by projecting the reference stimulus and the evaluation stimulus on a PC monitor that is 160 [mm] away from the position where the subject is sitting, and observing the reference stimulus and the evaluation stimulus. At this time, the periphery of the subject is set as a dark room environment by shielding the periphery of the subject with a dark screen.

FIG. 9 is a diagram illustrating an example of an evaluation result by the magnitude estimation method when the size and chromaticity of the target region 72 are varied. FIG. 9A illustrates a case where the size of the target region 72 is varied. B) shows a case where the Euclidean distance D indicating the color difference is varied. The experimental result of FIG. 9A shows that the color 2 indicated by the coordinates of the xy chromaticity diagram in FIG. 11 is the color of the background region 70 and the color of the target region 72 is the color 1, color 3 to color 6. Is the result of In FIG. 11, Y represents luminance [cd / m ² ].
From the results shown in the figure, it can be seen that, regardless of the color of the target region 72, the evaluation score tends to increase as the size of the target region 72 increases. This tendency was the same even when the color of the background region 70 was other colors 1 to 7 other than the color 2.

  In FIG. 9B, the color difference between the target area 72 and the background area 70 is expressed by the Euclidean distance D, and the Euclidean distance D is plotted for each size of the target area with the horizontal axis and the average value of the evaluation points as the vertical axis. Is. From this figure, it is recognized that the evaluation score tends to increase as the Euclidean distance D increases. For example, when the background area 70 is color 1 and the target area 72 is color 2, and when the background area 70 is color 2 and the target area 72 is color 1, the color between the background area 70 and the target area 72 is changed. When they are replaced, the Euclidean distance D is the same. For this reason, in the same figure, even if the Euclidean distance D and the target size are the same, a plurality of evaluation points are plotted for each Euclidean distance D.

FIG. 10 is a diagram illustrating an example of an evaluation result by the magnitude estimation method when the size and luminance ratio of the target region 72 are varied. FIG. 10A illustrates a case where the size of the target region 72 is varied. B) shows a case where the luminance ratio is varied.
From the results shown in the figure, it can be seen that the evaluation point increases as the size of the target region 72 increases or the luminance ratio increases.

The multiple regression coefficient α for the Euclidean distance contrast amount Q and the multiple regression coefficient β for luminance are derived by performing multiple regression analysis on the evaluation results obtained by the magnitude estimation method.
That is, when deriving the multiple regression coefficient α of the Euclidean distance contrast amount Q, in the magnitude estimation method, the evaluation stimulus given to the subject at the t-th time is DQt, and the average score of the evaluation points for this evaluation stimulus DQt is CQt. Then, in the same manner as the color difference conspicuous evaluation unit 29, u ′ image and v ′ image are respectively generated, Δu′v ′ color difference image is generated, and this is wavelet decomposed to level 9 to obtain nine images (details). Image). Then, the multiple regression analysis is executed with the value of the pixel at the center of the detail image (Euclidean distance contrast amount) as the explanatory variable and the evaluation average point CQt as the explained variable. As a result, nine multiple regression coefficients α1 to α9 that reflect human perception for each level are obtained. These levels indicate a range (size) in which the same color difference (or the same Euclidean distance D or the same pixel value) continues, and the size and the Euclidean distance contrast amount Q are conspicuous with the multiple regression coefficient α. It can be said.

  In the case of deriving the multiple regression coefficient β of luminance, in the magnitude estimation method for luminance, DLt is the evaluation stimulus given to the subject for the tth time, and CLt is the average score of evaluation points for this evaluation stimulus DLt. Here, a negative sign is added to the evaluation average point CLt for the evaluation stimulus DLt in which the target area luminance / background area luminance is 1.00 or less. By adding a negative sign to the evaluation average point CLt, the target area luminance / background area luminance is 1.00 or more, the stimulus (positive contrast stimulus) and the conspicuousness (positive contrast conspicuousness) caused by the contrast (positive contrast), the target The difference between the stimulus (reverse contrast stimulus) in which the region brightness / background region brightness is 1.00 or less and the conspicuousness (reverse contrast conspicuousness) caused by the contrast (reverse contrast conspicuousness) is expressed by a sign.

  Then, a logarithmic luminance distribution image of the evaluation stimulus DLt is generated in the same manner as the logarithmic luminance distribution image generation unit 45 of the luminance conspicuous evaluation unit 30, and the logarithmic luminance distribution image is set to level 9 in the same manner as the Wavelet decomposition unit 47. Wavelet decomposition until a total of nine images (detailed images) are obtained. Then, the multiple regression analysis is executed with the value of the center pixel of the detail image (logarithmic luminance contrast amount) as the explanatory variable and the evaluation average point CLt as the explanatory variable. Thereby, for each level, nine multiple regression coefficients from β1 to β9 reflecting human perception are obtained. Note that these levels indicate a range (size) in which the same luminance continues as described above, and it can be said that the size, luminance contrast amount L, and conspicuousness are associated by the multiple regression coefficient β.

Next, the conspicuous image generation processing by the conspicuous image generation device 11 will be described.
FIG. 12 is a flowchart of the conspicuous image generation process.
When the captured image data 5 is input (step S1), the conspicuous image generation device 11 individually evaluates conspicuous color difference and conspicuous luminance.
In other words, in the color difference conspicuous evaluation, each of the u ′ image generating unit 31 and the v ′ image generating unit 33 of the color difference conspicuous evaluating unit 29 generates a u ′ image and a v ′ image based on the captured image data 5 (steps). S2, S3). Next, the first Wavelet decomposition unit 35 Wavelet decomposes each of the u ′ image and the v ′ image to 9 levels (steps S4 and S5), and the Δu ′ intermediate image generation unit 36 and the Δv ′ intermediate image generation unit 37 approximate. The Wavelet is re-synthesized excluding the image (bias) to generate a Δu ′ intermediate image and a Δv ′ intermediate image, respectively (steps S6 and S7). Next, the Δu′v ′ color difference image generation unit 38 calculates the color difference of each pixel at the Euclidean distance D based on the Δu ′ intermediate image and each pixel Δu ′, Δv ′ of the Δv ′ intermediate image and the equation (1). The displayed Δu′v ′ color difference image 81 is generated (step S8).

Next, the second Wavelet decomposition unit 39 generates a 9-level Wavelet decomposition image by Wavelet analysis on the Δu′v ′ color difference image 81 (step S9).
In the Wavelet decomposition performed by the first Wavelet decomposition unit 35 and the second Wavelet decomposition unit 39, for example, a substantially symmetrical function such as symlet6 is used as an orthogonal Wavelet, and the u ′ image and the v ′ image shown in FIG. For each Δu′v ′ color difference image 81 shown in FIG. 5C, wavelet decomposition is performed up to level N (N = 9 in this embodiment) while increasing the level one by one.
For example, in the second wavelet decomposition unit 39, in the level 1 wavelet decomposition, as shown in FIG. 13, the Δu′v ′ color difference image 81 is converted into LL (Lv1) (L: Low-pass components), HL (Lv1) (H : High-pass components), LH (Lv1), and HH (Lv1) subband images (detailed images) 83LL, 83HL, 83LH, and 83HH.
The subband image 83LL of LL (Lv1) is obtained by extracting a low frequency component of the change in the Euclidean distance D in the Δu′v ′ color difference image 81, and the image size of the Δu′v ′ color difference image 81 is ¼ times. Corresponds to a coarse image that has been shrunk.
In addition, subband images 83HL, 83LH, and 83HH of HL (Lv1), LH (Lv1), and HH (Lv1) are high-frequency components in the vertical direction among changes in Euclidean distance D in Δu′v ′ color difference image 81, respectively. The high-frequency component in the horizontal direction and the high-frequency component in the oblique direction are extracted. In these subband images 83HL, 83LH, 83HH, the value of each pixel represents the Euclidean distance contrast amount Q1 at level 1 in the vertical, horizontal, and diagonal directions for each pixel of the Δu′v ′ color difference image 81. Giving.
Further, for example, as shown in FIG. 14A, when pixels of the Euclidean distance D are distributed in a donut shape (radius <R), the target pixel O at the center of the donut shape and the target pixel Although there is no contrast of the Euclidean distance D with the surrounding pixel P outside the donut shape, which is away from the radius R from O, the contrast occurs due to the donut shape in the range including the surrounding pixel P from the target pixel O. Yes. Even in this case, according to Wavelet decomposition using a function such as symlet6, the Euclidean distance contrast amount Q after decomposition does not simply become zero. As shown in FIG. Since the Euclidean distance contrast amount Q between O and the surrounding pixels P is calculated, a considerable amount of contrast is calculated at a certain level.
In addition to the Euclidean distance D, for example, when signal values of chromaticity components and luminance are distributed in a donut shape (where radius <R), according to Wavelet decomposition using a function such as symlet 6, it corresponds to a certain level. A quantity contrast is calculated.

In the next level 2 wavelet decomposition, as shown in FIG. 13, wavelet decomposition is performed on the subband image 83LL of LL (Lv1), and LL (Lv2), HL (Lv2), LH (Lv2), and HH ( Lv2) are subdivided into four subband images 83LL, 83HL, 83LH, and 83HH.
Thereafter, in the same manner, the Wavelet decomposition is performed on only the subband image 83LL of LL (Lvn (n = 1 to N)) one after another until the level N is reached while increasing the level one by one. Sub-band images 83LL, 83HL, 83LH, and 83HH are generated for each level.

  At this time, by repeating Wavelet decomposition on the subband image 83LL of LL (Lvn (n = 1 to N)), the Δu′v ′ color difference image 81 is sequentially increased to 1/4 times each time Wavelet decomposition is performed. A coarse sub-band image 83LL that has been reduced is generated. As a result, the higher the level of Wavelet decomposition, the larger the image relative to the Wavelet function. For each pixel of the Δu′v ′ color difference image 81, the higher the level, the farther the surrounding pixels. A comparison of the Euclidean distance D with P is obtained.

Returning to FIG. 12, when Wavelet decomposition up to the N level is performed in step S <b> 9, the coefficient processing unit 40 performs horizontal, vertical, and diagonal directions at each level for all pixels of the Δu′v ′ color difference image 81. The intermediate generation values IQ1 to IQN of the Euclidean distance contrast amount Q of each level are obtained for all the pixels of the Δu′v ′ color difference image 81 based on the Euclidean distance contrast amount Q and the multiple regression coefficient α corresponding to the level. (Step S10).
For example, when subband images 83LL, 83HL, 83LH, and 83HH obtained by wavelet decomposition at level 3 are targeted, as shown in FIG. 15, all pixel values of the subband image 83LL of LL (Lv3) are expressed as “ In addition, the multiple regression coefficients α3 corresponding to level 3 are respectively multiplied to all the pixels of HL (Lv3), LH (Lv3), and HH (Lv3) indicated by diagonal lines.
Then, the level 3 sub-band images 83LL, 83HL, 83LH, and 83HH are wavelet-combined by orthogonal wavelets (for example, symbolet6) used in the wavelet decomposition.
Thereafter, wavelet synthesis is repeated while decreasing the level one by one to the original image level (0 level). At this time, the values of all the pixels of the subband images 83HL, 83LH, and 83HH at each level are synthesized as “0”. Is repeated.
As a result, a color difference noticeable image 85 corresponding to level 3 in which the intermediate generation value IQ3 for the Euclidean distance contrast amount Q corresponding to level 3 is stored in all pixels of the Δu′v ′ color difference image 81 is obtained.
Similarly, color difference noticeable images corresponding to the respective levels are generated for levels 1 to N, and the color difference noticeable images corresponding to the respective levels are synthesized (adding values of corresponding pixels of the respective noticeable images). Thus, data of the color difference noticeable image 85 in which the intermediate generation value IQ is stored in all the pixels of the Δu′v ′ color difference image 81 is generated (step S11).

Next, the luminance noticeability evaluation will be described. The logarithmic brightness distribution image generation unit 45 of the brightness noticeability evaluation unit 30 generates a logarithmic brightness distribution image based on the captured image data 5 (step S12). Next, the Wavelet decomposition unit 47 performs Wavelet decomposition of the logarithmic luminance distribution image to level 9 using the symbol 6 in the same manner as the second Wavelet decomposition unit 39 of the color difference conspicuous evaluation unit 29, and calculates the luminance contrast amount L at each level. Obtained (step S13).
Next, the coefficient processing unit 49 multiplies the sub-band image 83 at each level by the multiple regression coefficients β1 to β9 corresponding to the level to obtain intermediate generation values IL1 to IL9, and the Wavelet synthesis unit 51 selects each level. The subband image 83 is wavelet synthesized (step S14), and a luminance noticeable image in which the luminance noticeable prediction value ML is stored in all the pixels of the logarithmic luminance distribution image is generated (step S15).

Next, the luminance conspicuous prediction value code conversion unit 65 of the imaging unit 25 generates a luminance conspicuous image in which the luminance conspicuous prediction values ML of the negative signs are all positive in the luminance conspicuous image (step S16). A code matrix of the luminance conspicuous image indicating the sign of each pixel of the previous luminance conspicuous image is generated (step S17).
Thereafter, the color difference / luminance conspicuous composition unit 67 adds the sign-converted luminance conspicuous prediction value ML and the color difference conspicuous prediction value MQ to calculate the conspicuous value U, and the multistage evaluation value conversion unit 69 calculates the conspicuous value U. Is converted into a 13-step evaluation value to generate a conspicuous image (step S18). Next, in order to clearly show the difference between the luminance contrast and the contrast, the evaluation value code conversion unit 71 adds the code matrix generated in step S17 to the image, and the pixel whose luminance conspicuous predicted value ML is a negative sign. A negative sign is added to the 13-stage evaluation value (step S19). As a result, a conspicuous image in which multi-stage evaluation values from “+13” to “−13” are stored in all pixels of the captured image is generated, and data of the conspicuous image is output from the image output unit 27 (step S20). ). In this conspicuous image, each evaluation value is color-coded into a pseudo color. Thereby, it becomes possible to quantitatively grasp the conspicuousness of the evaluation object captured in the captured image.

FIG. 16 is a sample diagram simulating the captured image data 5 obtained by photographing the light environment 3 in which the disk 92 is arranged on the background 91 of one color, and shows the size and chromaticity of the disk 92 variably. It is a figure which shows the luminance distribution of the captured image data. FIG. 17 is a sample diagram of conspicuous images in which the conspicuousness of each captured image data 5 is evaluated.
In these drawings, the disk 92 corresponds to the target pixel O to be evaluated, and the background 91 corresponds to the surrounding pixel P. As shown in these figures, when the color difference between the disc 92 and the background 91 is constant, it is evaluated that the disc 92 is more conspicuous as the size of the disc 92 increases, and when the size of the disc 92 is the same. It can be seen that the higher the color difference between the disk 92 and the background 91, the more it is evaluated to stand out. That is, according to this conspicuous evaluation, it is understood that the conspicuousness considering both the color difference between the disc 92 and the background 91 to be evaluated and the size of the disc 92 is evaluated in an integrated manner.

As described above, according to the present embodiment, when there is no color difference between the evaluation target and the surroundings, the Euclidean distance that is the contrast amount between the evaluation target and the surroundings is evaluated with respect to the captured image obtained by capturing the evaluation target. When the contrast amount Q is large, the evaluation object is highly evaluated for its conspicuousness according to the degree of its size, and based on this evaluation, a color difference conspicuous image 85 is generated from the captured image.
With this configuration, the conspicuousness based on the color contrast between the evaluation object and the surroundings can be quantitatively evaluated from the color difference conspicuous image 85.

In particular, according to the present embodiment, the color difference between the object to be evaluated and the surrounding color is calculated as a Euclidean distance D between coordinates after obtaining a contrast amount for each coordinate component on the u′v ′ chromaticity diagram by wavelet analysis, and then wavelet. The Euclidean distance contrast amount Q indicating the color difference contrast amount is obtained by analysis, and the conspicuousness is evaluated based on the Euclidean distance contrast amount Q.
With this configuration, the captured image data 5 is divided into u ′ images and v ′ images for each chromaticity component, and compared with the case where the conspicuousness is evaluated by wavelet analysis for each, explanatory variables and multiple regression coefficients α that explain the conspicuousness of the color Therefore, the reliability per explanatory variable can be increased. Further, it is not necessary to convert a detail image including a negative value of the u ′ image and the v ′ image into a positive value. For this reason, the restoration property of the original image by Wavelet synthesis is not impaired, and the conspicuousness of the color to be evaluated included in the captured image data 5 can be accurately evaluated.

Further, according to the present embodiment, when the evaluation target and the surrounding luminance are equal to the captured image obtained by capturing the evaluation target, the evaluation is not noticeable. The conspicuousness of the evaluation object was highly evaluated according to the degree of the degree, and based on the evaluation results of the conspicuousness of the color difference and the conspicuousness of the brightness, a conspicuous image showing the distribution of the conspicuousness of the chromaticity and the brightness of the captured image was generated. Thereby, the conspicuousness due to the difference in chromaticity and luminance can be quantitatively evaluated.
In particular, in the prior art, when evaluating an evaluation object based on luminance, the conspicuousness is evaluated based on the brightness. Therefore, the darker the evaluation object is, the less conspicuous it is. However, in reality, when the periphery of the evaluation target is bright, the darker the evaluation target, the more conspicuous the contrast becomes. Thus, in the conventional technique, an evaluation result different from the actual result may be obtained. Further, even when the evaluation target is bright, if the surroundings are equally bright, the evaluation target becomes inconspicuous, but an evaluation result that is conspicuous in the conventional technique is obtained. According to this embodiment with respect to such a conventional technique, the conspicuousness of the brightness is evaluated based on the brightness contrast with the surroundings, so that the conspicuousness of the evaluation target can be accurately evaluated.

  Further, in the present embodiment, the conspicuousness due to the inverse contrast of luminance is indicated by a negative sign, and the conspicuousness due to the direct contrast of luminance is indicated by a positive sign, so that these are separately imaged into conspicuous images. Thereby, it can be discriminated whether the evaluation object is conspicuous by the direct contrast or the reverse contrast.

  The above-described embodiments merely show one aspect of the present invention, and can be arbitrarily modified and applied within the scope of the present invention.

For example, in the above-described embodiment, the Euclidean distance D on the u′v ′ chromaticity diagram is used for the color difference between the color to be evaluated and the surrounding color. However, other equivalent colors such as the uv chromaticity diagram are used. A degree diagram may be used.
Further, coordinates in a uniform color space may be used instead of the uniform chromaticity diagram.
More specifically, the luminance distribution of the light environment 3 is usually complicated, and there are many cases where the luminance to be evaluated and the surrounding luminance are not the same, but this u'v 'chromaticity diagram shows the comparison of colors with different luminance. Not considered. That is, it is not guaranteed that the distance between two points on the chromaticity diagram is equal to the perceptual color difference.
Therefore, in the comparison of colors having different luminances, the coordinates in the uniform color space designed so that the distance between the two points in the color space is equal to the perceptual color difference are replaced with the u′v ′ chromaticity diagram. It is also possible to display the color conspicuously and improve the conspicuousness evaluation accuracy.

FIG. 18 is a flowchart of conspicuous image generation processing when the L ^* a ^* b ^* color system (JIS Z8729) is used as the uniform color space. When the L ^* a ^* b ^* color system is used as the uniform color space, the chromaticity coordinates (u ′, v ′) are the perceptual chromaticity indices a ^* and b ^*, and the luminance L is the perceptual lightness index L ^*. As follows.
That is, as shown in FIG. 18, after the captured image data 101 is input (step S101), in the color difference conspicuous evaluation, the color difference conspicuous evaluation unit 29 generates an a ^* image and a b ^* image (step S102). , S103), wavelet decomposed to 9 levels (steps S104, S105), and remove the approximate image (bias) and re-synthesize the wavelet to generate Δa ^* intermediate image and Δb ^* intermediate image (step S106). , S107). Next, the color difference conspicuous evaluation unit 29 determines the L ^* a ^* b ^* color of the target pixel O and the surrounding pixels P based on the values Δa ^* and Δb ^* of each pixel of the Δa ^* intermediate image and Δb ^* intermediate image. A Δa ^* b ^* color difference image in which the color difference in the system is displayed with the Euclidean distance D is generated (step S108).

Then, in order to evaluate the color conspicuousness based on the Δa ^* b ^* color difference image, the color difference conspicuous evaluation unit 29 generates a 9-level Wavelet decomposition image by wavelet analysis on the Δa ^* b ^* color difference image (step S109), the Euclidean distance of each level based on the Euclidean distance contrast amount Q in the horizontal, vertical and diagonal directions of each level and the multiple regression coefficient corresponding to that level for all pixels of the Δa ^* b ^* color difference image. Intermediate generation values IQ1 to IQN of the contrast amount Q are obtained (step S110).
Then, the color difference conspicuous evaluation unit 29 synthesizes the images of the respective levels (by adding the values of the corresponding pixels of each conspicuous image), thereby generating an intermediate generation value IQ for all the pixels of the Δa ^* b ^* color difference image. Is generated (step S111).

On the other hand, the luminance conspicuous evaluation unit 30 generates a lightness L ^* image based on the captured image data 5 (step S112), and after performing Wavelet decomposition to 9 levels in the same manner as in steps S105 and S106 (step S113), A ΔL ^* intermediate image is generated by re-synthesis of Wavelet, excluding the approximate image (bias) (step S114). Note that the processing in these steps S113 and S114 can be omitted.
Next, the luminance conspicuous evaluation unit 30 wavelet decomposes ΔL ^* intermediate image to level 9 using symlet 6 to obtain a brightness contrast amount corresponding to the luminance contrast amount at each level (step S115). After the intermediate generation values IL1 to IL9 are obtained by multiplying the subband image 83 by the multiple regression coefficients corresponding to the levels, the subband images 83 of these levels are wavelet synthesized (step S116), and the lightness L ^* image A lightness conspicuous image in which lightness conspicuous predicted values are stored in all the pixels is generated (step S117).

Then, the imaging unit 25 generates a lightness noticeable image in which all of the lightness noticeable prediction values of the negative signs are positive in the lightness noticeable image (step S117), and the sign of each pixel of the lightness noticeable image before code conversion is also determined. A code matrix indicating the code is generated (step S118).
Thereafter, the imaging unit 25 simply adds the sign-converted lightness conspicuous predicted value and the color difference conspicuous predicted value to calculate a conspicuous value, and converts this conspicuous value into a 13-level evaluation value to generate a conspicuous image (Step S119). Next, the imaging unit 25 adds the code matrix generated in step S118 to the conspicuous image to clarify the difference in conspicuity due to the positive contrast and reverse contrast of the lightness, and 13 of the pixels whose lightness conspicuous predicted value is a negative sign. A negative sign is added to the stage evaluation value (step S120). As a result, a conspicuous image is generated in which multistage evaluation values from “+13” to “−13” are stored in all pixels of the captured image, and data of the conspicuous image is output from the image output unit 27 (step S121). ).

In the conspicuous image processing shown in FIG. 18, when the L ^* a ^* b ^* color space is used, it is not necessary to separate the chromaticity and the lightness, and the perceptual lightness index L ^* is included in the perceptual chromaticity indices a ^* and b ^{*. A} color difference image may be generated by calculating the Euclidean distance D using three variables including ^* as variables.
FIG. 19 is a flowchart of such conspicuous image processing. In this figure, the same processes as those in FIG. 18 are denoted by the same reference numerals and description thereof is omitted.
As shown in this figure, in step S208, the color difference conspicuous evaluation unit 29 is based on Δa ^* intermediate image, Δb ^* intermediate image, and values Δa ^* , Δb ^* , ΔL ^* of each pixel of the ΔL ^* brightness difference image. Then, a ΔE ^* ab color difference image in which the color difference in the L ^* a ^* b ^* color space between the target pixel O and the surrounding pixel P is displayed with the Euclidean distance D is generated (step S208), and based on this ΔE ^* ab color difference image Then, color noticeability is evaluated using Wavelet analysis.
Since the conspicuousness of brightness is already reflected in the color conspicuous image generated based on this ΔE ^* ab color difference image, the lightness conspicuous image of only positive values (step S118 in FIG. 18), and positive by simple addition. The conspicuous image generation of only the values (step S120 in FIG. 18) is not necessary.

Although the L ^* a ^* b ^* color system is illustrated as the uniform color space, other color systems classified into the uniform color space such as the L ^* u ^* v ^* color system may be used. Of course.

Further, based on the coordinates of the uniform color space, saturation indicating vividness, hue indicating hue, and brightness indicating brightness may be calculated, and the conspicuity may be evaluated using these Euclidean distances as color differences.
FIG. 20 is a flowchart of such conspicuous image processing. In this figure, the same processes as those in FIGS. 18 and 19 are denoted by the same reference numerals, and the description thereof is omitted.
As shown in this figure, when captured image data is input, the color difference conspicuous evaluation unit 29 generates a C ^* saturation image and an H ^* hue image based on the coordinates of the uniform color space (steps S302 and S303). ), Wavelet decomposition is performed (steps S304 and S305), and the wavelet is re-synthesized by removing the approximate image (bias) to generate a ΔC ^* saturation difference image and a ΔH ^* hue difference image (steps S306 and S307). .
Then, based on the values ΔC ^* , ΔH ^* , ΔL ^* of each pixel of the ΔC ^* saturation difference image, ΔH ^* hue difference image, and ΔL ^* brightness difference image, the saturation and hue obtained from the uniform color space. Then, a ΔE ^* color difference image in which the color difference in the three-dimensional space of brightness is displayed with the Euclidean distance D is generated (step S308), and color conspicuity is evaluated using Wavelet analysis based on this ΔE ^* color difference image. .

  In the above-described embodiment, the case where the luminance contrast amount L is obtained for each pixel of the logarithmic luminance distribution image using the wavelet transform has been exemplified. However, the present invention is not limited to this. Even if a plurality of N filters having different ranges (number of pixels) according to each level are used, the luminance contrast amount L for each level can be obtained.

As an application example of the conspicuous evaluation system 1 of the present invention, it is evaluated whether an object that is desired to be conspicuous with respect to surroundings such as advertisements, signs, road traffic signs, and gaze guidance lights is conspicuous in the installation environment. It becomes possible to do.
In addition, for example, by evaluating the conspicuousness of liquid crystal display when viewed from the user, it can be used to determine the color unevenness of the liquid crystal display acceptable to the user, and it is possible to prevent excessive quality It is.

  The above-mentioned conspicuous image generation program 15 may be recorded on a computer-readable recording medium such as a CD or DVD, and may be distributed via an electric communication line such as the Internet. Is also possible.

DESCRIPTION OF SYMBOLS 1 Evaluation system 3 Light environment 5 Captured image data 9 Conspicuous image data 11 Conspicuous image generation apparatus 15 Conspicuous image generation program 23 Conspicuous evaluation part 25 Imaging part 29 Color difference conspicuous evaluation part 30 Luminance conspicuous evaluation part 35 1st Wavelet decomposition part 38 (DELTA) u ' v ′ color difference image generation unit 39 second wavelet decomposition unit 40, 49 coefficient processing unit 47 wavelet decomposition unit 43 addition unit 67 color difference / luminance conspicuous synthesis unit 69 conspicuous image generation unit 81 Δu′v ′ color difference image 83 subband image 85 conspicuous image L Luminance contrast amount Q Euclidean distance contrast amount (color difference contrast amount)
P Surrounding pixels U Conspicuous value ML Luminance conspicuous predicted value MQ Color difference conspicuous predicted value α, β Multiple regression coefficient

Claims (4)

Converts the value of each pixel of the captured image obtained by imaging the evaluation target into a uniform chromaticity diagram designed so that the Euclidean distance between two points corresponds to a perceptual color difference, or coordinates in a uniform color space , it was evaluated that there is no noticeable when the color difference based on the Euclidean distance of the evaluated coordinates and surrounding coordinate is not, the coordinates and color differences contrast amount based on the Euclidean distance of the periphery of the coordinates of the object of evaluation in the case of a large A color difference conspicuity evaluation unit that highly evaluates the conspicuity of the evaluation object according to the degree of the size;
A conspicuous image generating apparatus comprising: an imaging unit that generates a conspicuous image indicating a conspicuous distribution due to the color difference of the captured image based on an evaluation result of the color difference conspicuous evaluation unit. The color difference conspicuous evaluation unit is
After the color difference of the evaluation object and the surroundings, was calculated as the Euclidean distance between equal chromaticity diagram, or seeking contrast amount of each coordinate component of the coordinate of the uniform color space coordinates by Wavelet analysis, the color difference by Wavelet Analysis seeking Euclidean distance contrast amount indicating the contrast amount, the image generating apparatus conspicuous according to claim 1, characterized in that to evaluate the conspicuous based on the Euclidean distance contrast amount. When the evaluation object and the surrounding luminance are equal to the captured image obtained by imaging the evaluation object, the evaluation is not noticeable, and the degree of size when the evaluation object and the surrounding luminance contrast amount are large And a brightness conspicuous evaluation unit that highly evaluates the conspicuousness of the evaluation object according to
The imaging unit
The conspicuous image according to claim 1 or 2, wherein a conspicuous image showing a distribution of conspicuous color difference and luminance of the captured image is generated based on the evaluation results of the color difference conspicuous evaluation unit and the luminance conspicuous evaluation unit. Image generation device. Computer
Converts the value of each pixel of the captured image obtained by imaging the evaluation target into a uniform chromaticity diagram designed so that the Euclidean distance between two points corresponds to a perceptual color difference, or coordinates in a uniform color space , it was evaluated that there is no noticeable when the color difference based on the Euclidean distance of the evaluated coordinates and surrounding coordinate is not, the coordinates and color differences contrast amount based on the Euclidean distance of the periphery of the coordinates of the object of evaluation in the case of a large A means for highly evaluating the conspicuousness of the evaluation object according to the degree of the size, and
A conspicuous image generation program that functions as means for generating a conspicuous image indicating a conspicuous distribution due to color difference of the captured image based on an evaluation result of the color difference conspicuous evaluation unit.