JP2002245464A - Device and method for image evaluation, and program making computer evaluate image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To output an evaluated value which is correlated to the subjective evaluation of an image. SOLUTION: An image evaluating device 100 is equipped with an image data input part 101 which inputs spatial optical information of the image, a convolutional arithmetic part 103 which performs convolutional arithmetic between a spatial arithmetic filter stored in an arithmetic filter storage part 106 and a spatial arithmetic filter for the spatial optical information of the image inputted by the image data input part 101, and an image quality evaluation part 105 which predicts the image quality that a person feels according to the output of the convolutional arithmetic part 103.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image evaluation device for evaluating the quality of a hard copy image, an image evaluation method, and a program for causing a computer to evaluate an image. More particularly, the present invention relates to a logistic regression analysis based on a multi-channel model. The present invention relates to an image evaluation device, an image evaluation method, and a program for causing a computer to evaluate an image by using the evaluation formula for which the coefficient has been determined.

[0002]

2. Description of the Related Art It has been known that a plurality of channels having spatial frequency selectivity (hereinafter, appropriately referred to as multi-channels) exist in human visual functions. According to neurophysiological and anatomical studies, retinal ganglion cells have a substantially circularly symmetric receptive field.
It has been clarified that there are on-centre cells having an excitatory region in the central part of the receptive field and an inhibitory region in the peripheral part, and off-centre cells having the opposite characteristics. It is thought that the on-centered cell and the off-centered cell form the visual pathway of a spatial frequency selective channel. FIG. 17 is an explanatory view showing the schematic shape of the receptive field of the channel, and is a diagram in which the output from each photoreceptor is plotted on the plane of the receptive field in the z-axis direction and displayed from the horizontal direction. The figure shows on-centre cells.

On the other hand, as a method of evaluating the image quality of a hard copy image, there is a technique disclosed in Japanese Patent Application Laid-Open No. 7-325922. That is, in order to take into account human visual characteristics, image quality is converted by using the amount obtained by multiplying the MTF characteristic (VTF) of human vision with image information converted into a spatial frequency domain (power spectrum), which is a physical quantity. Is used.

[0004]

However, the prior art has the following problems. That is, the conventional image quality evaluation method has a problem that it is difficult to integrate the perception of texture and the granularity of a digital image. In addition, there is a problem that it is difficult to detect an edge using only a texture.

[0005] This is considered to be due to the fact that the spatial frequency characteristic is only approximately effective because the human visual system that perceives image quality is a very nonlinear system. In addition, since the visual MTF characteristic is obtained based on the data of the threshold value of the contrast sensitivity to a human sine wave-like stimulus, the visual MTF characteristic is merely one aspect of the visual phenomenon by multi-channel. Conceivable.

For example, when two images having the same density are viewed side by side, a human perceives that the appearance of both images is different due to the difference in texture. On the other hand, the conventional method using the spatial frequency characteristic has a problem that the same appearance is evaluated as being evaluated.

In other words, there is a problem that an evaluation value correlated with the subjective evaluation of an image may not be output.

The present invention has been made in view of the above, and has as its object to output an evaluation value correlated with the subjective evaluation of an image.

[0009]

According to another aspect of the present invention, there is provided an image evaluation apparatus, comprising: an image input unit for inputting spatial and optical information of an image; A spatial operation filter used for the obtained spatial optical information, a spatial operation filter, and a convolution operation unit for performing a convolution operation of the spatial operation filter on the spatial optical information of the image input by the image input unit. And image quality prediction means for predicting the image quality perceived by humans based on the output of the convolution operation means. That is, the invention according to claim 1 performs image quality evaluation based on characteristics of a human visual system.

[0010] The image evaluation apparatus according to claim 2 is
2. The image evaluation device according to claim 1, wherein the spatial operation filter is an isotropic differential operation filter. That is, the invention according to claim 2 quantifies an image quality factor caused by a two-dimensional image structure based on characteristics of a human visual system.

Further, the image evaluation device according to the third aspect is characterized in that:
3. The image evaluation device according to claim 1, wherein the spatial operation filter includes a plurality of filters having different characteristics, and the image quality prediction unit outputs from the convolution operation unit corresponding to each of the plurality of filters. It is characterized by predicting the image quality perceived by humans using everything. That is, the invention according to claim 3 quantifies the image quality factors caused by the image structures of various scales based on the characteristics of the human visual system, and quantifies the image quality comprehensively.

[0012] The image evaluation device according to claim 4 is
4. The image evaluation device according to claim 3, wherein the plurality of filters are optimized for spatial frequency characteristics of a human visual system. That is, the invention according to claim 4 is based on the spatial frequency characteristics of the human visual system.
Perform more precise image quality quantification.

[0013] The image evaluation apparatus according to claim 5 is
5. The image evaluation device according to claim 1, wherein the spatial calculation filters have different characteristics. 5.
It is characterized by having a number of filters. That is,
The invention according to claim 5 performs more precise quantification of image quality based on a multi-channel model of the human visual system.

Further, the image evaluation device according to claim 6 is
The image evaluation device according to claim 3, 4 or 5,
The image processing apparatus further includes a filter selection unit that selects a required number of the plurality of filters. That is, the invention according to claim 6 provides a perception corresponding to a plurality of types of image defects in which image structures are distributed on different visual scales based on a multi-channel model of a human visual system.
Quantify faster and more precisely.

Further, the image evaluation device according to claim 7 is
7. The image evaluation apparatus according to claim 1, further comprising: a filter size changing unit configured to change a size of the spatial operation filter. That is, the invention according to claim 7 performs image quality evaluation independent of the image density of the evaluated image. This is because, in order to perform image quality evaluation based on the multi-channel model, it is desirable that the image density of the evaluated image be approximately equal to the image density on the retina.

Further, the image evaluation device according to the present invention is characterized in that:
Image data input means for inputting image data of an image to be evaluated, spatial calculation filter selection means for selecting a spatial calculation filter to be used for the image data input by the image data input means, and the image data input means A convolution operation unit that performs a convolution operation on the image data input by using the spatial operation filter selected by the spatial operation filter selection unit, based on a result of the convolution operation performed by the convolution operation unit. Image quality evaluation means for evaluating the quality of the image. That is, the invention according to claim 8 performs image quality evaluation based on characteristics of a human visual system.

Further, the image evaluation apparatus according to claim 9 is
9. The image evaluation device according to claim 8, wherein the spatial operation filter is an isotropic differential operation filter. That is, the invention according to claim 9 quantifies an image quality factor caused by a two-dimensional image structure based on characteristics of a human visual system.

According to a tenth aspect of the present invention, in the image evaluation apparatus of the ninth aspect, the isotropic differential operation filter is provided with a function LG (r) = (− 1 / (π
σ ⁴ )) · (1−r ² / (2σ ² )) · exp (−r ² /
(2σ ² )) (where r is r = sqrt (x ² + y ² )
Where x and y are coordinates in a coordinate system centered on a certain position of the image data of the image to be evaluated, and σ corresponds to five types of channels in a multi-channel model of a visual system. This is a characteristic constant). That is, the invention according to claim 10 is
Quantify finer image quality based on multi-channel model of human visual system.

An image evaluation apparatus according to an eleventh aspect is an image evaluation apparatus for comparing two images and outputting an evaluation according to a subjective evaluation of whether the two images look the same or appear to have a difference. An image data input unit for inputting image data of two images to be evaluated; and an absolute value ΔT of a difference between texture amounts and an average value Tmean of texture amounts based on the data input by the image data input unit. Texture amount calculating means for calculating the halftone dot area ratio of the two images to be evaluated, and a halftone dot area ratio obtaining means for obtaining an average value Mmean of the halftone dot area ratio,
The absolute value ΔT of the difference between the texture amounts calculated by the texture amount calculating means and the average value Tmea of the texture amounts
n, the absolute value ΔM of the difference between the halftone dot area ratios obtained by the halftone dot area ratio obtaining means and the average value Mm of the halftone dot area ratios
and an image evaluation value calculating means for calculating an image evaluation value D according to equation (1) using E.ean (where p _{1 to} p ₉ are based on a subjective evaluation experiment conducted in advance. The determined constant). That is, the invention according to claim 11 calculates a difference in appearance due to the texture of an image having the same density macroscopically.

According to a twelfth aspect of the present invention, in the image evaluation apparatus according to the eleventh aspect, the image evaluation value D calculated by the image evaluation value calculation means is further expressed by the following equation (2). An image evaluation value correction unit for correcting the image evaluation value PD based on the image evaluation value PD is provided (however, q1 to q4 are constants determined by regression analysis). That is, the invention according to claim 12 outputs an evaluation value highly correlated with the subjective evaluation of the image.

According to a thirteenth aspect of the present invention, in the image evaluation apparatus according to the first aspect, the input density of the image data input by the image data input means is determined by a person. Input density adjusting means for adjusting the image density on the retina at the time of observation so as to be substantially equal to the density is provided. That is, in the invention according to claim 13, the image density of the evaluated image is made substantially equal to the image density on the retina.

According to a fourteenth aspect of the present invention, there is provided an image evaluation method for predicting an image quality perceived by a human based on an output obtained by convolving a spatial and optical information of an image with a spatial operation filter. It is characterized by. That is, claim 1
The invention according to 4 performs image quality evaluation based on characteristics of a human visual system.

According to a fifteenth aspect of the present invention, in the image evaluation method of the fourteenth aspect, the spatial operation filter is an isotropic differential operation filter. That is, the invention according to claim 15 quantifies the image quality factor caused by the two-dimensional image structure based on the characteristics of the human visual system.

The image evaluation method according to claim 16 is the image evaluation method according to claim 14 or 15, wherein a plurality of spatial operation filter groups having different characteristics are used as the spatial operation filters. It is characterized in that the image quality perceived by a human is predicted by using all of the outputs obtained by convolution by the filter group. That is, the invention according to claim 16 quantifies the image quality factors caused by the image structures of various scales based on the characteristics of the human visual system, and comprehensively quantifies the image quality.

According to a seventeenth aspect of the present invention, in the image evaluation method of the sixteenth aspect, the spatial operation filter group is optimized for a spatial frequency characteristic of a human visual system. It is characterized by. That is, the invention according to claim 17 performs more precise quantification of image quality based on the spatial frequency characteristics of the human visual system.

The image evaluation method according to claim 18 is the image evaluation method according to any one of claims 14 to 17, wherein five spatial operation filters having different characteristics are used as the spatial operation filters. Is used. That is, the invention according to claim 18 performs more precise quantification of image quality based on a multi-channel model of the human visual system.

According to a nineteenth aspect of the present invention, in the image evaluation method of the sixteenth, seventeenth, or eighteenth aspect, a necessary number of the spatial operation filters are selected and used. And That is, claim 1
The invention according to 9 quantifies perception corresponding to a plurality of types of image defects in which image structures are distributed on different visual scales faster and more precisely based on a multi-channel model of the human visual system. .

According to a twentieth aspect of the present invention, in the image evaluation method according to any one of the fourteenth to nineteenth aspects, the size of the spatial operation filter is variable. . That is, the invention according to claim 7 provides an image evaluation device that performs image quality evaluation independent of the image density of the evaluated image. This is because, in order to perform image quality evaluation based on the multi-channel model, it is desirable that the image density of the evaluated image be approximately equal to the image density on the retina.

Further, the program according to claim 21 is:
A program for causing a computer to execute image evaluation, wherein the program causes a computer to function as each unit according to any one of claims 1 to 13. That is, the invention according to claim 21 is based on claims 1 to
Since the computer is a program that causes a computer to execute the means described in any one of the above aspects, it is possible to output an evaluation value that is correlated with the subjective evaluation of the image using the computer.

[0030]

Embodiments of the present invention will be described below in detail with reference to the drawings. Embodiment 1 FIG. In the first embodiment, first, an outline of a model of a visual system will be described. Next, an example in which the image evaluation apparatus is applied to a personal computer (hereinafter, appropriately referred to as a PC) will be described. Will be described.

(About the Model of the Visual System) Here, the multichannel theory of vision and the receptive field will be outlined.
Since Campbell and Robson concluded the existence of a channel with spatial frequency selectivity in human visual function based on the results of psychophysical experiments in 1968, various channels with this spatial frequency selectivity have been Psychophysical studies have been performed.

On the other hand, retinal ganglion cells have a substantially circularly symmetric receptive field, an on-center type cell having an excitatory region in the central portion of the receptive field and an inhibitory region in the peripheral portion, and the opposite off-center type cell. The presence of type cells has been demonstrated by neurophysiological and anatomical studies and is thought to form the visual pathway of spatially frequency-selective channels.

In addition, Wilson et al. Built a quantitative model of the multi-channel mechanism in humans based on data in the visual detection zone. Wilson et al. Described the receptive field classified into four types corresponding to these channels as the difference between two Gaussian functions (DOG: Diffe
The probability of detecting a stimulus in each channel is determined by the law of spatial probability addition (Low of probability su).
MMation), 0.25 to 16.0
It has been proved that the psychophysical data concerning the threshold detection of a one-dimensional spatial pattern in the range of [cycle / deg] can be basically explained.

That is, Wilson et al. Proposed four types of one-dimensional channels (multi-channels) existing at each point of the visual field. These were named N, S, T, U in ascending order of the size of the central area of the receptive field.
The size of these channels increases linearly with eccentricity, the distance from the fovea expressed in viewing angles. One-dimensionally, a channel having such a receptive field is considered to play a role of a detector for a change in the intensity of external light reflected on the retina.

Further, this quantitative model is extended and reinforced two-dimensionally by Mar et al. Where Mar
Predicts the existence of the smallest channel corresponding to midget ganglion cells in which the central part of the receptive field is formed by a single cone in developing Wilson et al.'S multi-channel theory in two dimensions. The model was built.

Assume that the best filter for detecting the intensity distribution of a stimulus is a ΔG filter. Here, Δ is a Laplace operator, and G is a two-dimensional Gaussian distribution having a standard deviation σ (where σ takes five unique values for each channel). G and ΔG are expressed by the following equations (3) and (4).
Is represented by

(Equation 3)

(Equation 4)

In the following, an image evaluation apparatus, an image evaluation method, and a program for causing a computer to evaluate an image, which are backed by a multi-channel model, by an isotropic differential operation filter using a G or ΔG function profile by Mar et al. Will be described. I do.

(Regarding Image Evaluation Apparatus) FIG. 1 is a functional block diagram when the image evaluation apparatus of the first embodiment is applied to a PC, and FIG. 2 is a hardware configuration of the image evaluation apparatus of the first embodiment. FIG. 4 is an explanatory diagram showing an example of the embodiment.

The PC 100 includes an image data input unit 101 for inputting image data of an image to be evaluated, and an operation filter selection unit for selecting a spatial operation filter to be used for the image data input by the image data input unit 101. 102, a convolution operation unit 103 that performs a convolution operation on the image data input by the image data input unit 101 using the spatial operation filter selected by the operation filter selection unit 102, and a convolution operation unit 103. And an image quality evaluation unit 104 that outputs an evaluation value of the quality of the image to be evaluated based on the result of the convolution operation. In the following, the PC 100 may be referred to as the image evaluation device 100.

The image quality evaluation unit 104 obtains a dot area ratio of an image to be evaluated.
The evaluation value of the quality of the image is output using the output value from 5. Depending on the mode of use, a physical attribute of the image other than the dot area ratio, for example, a reflectance may be used. It is assumed that the operation filter selected by the operation filter selection unit 102 is stored in the operation filter storage unit 106. The operation filter storage unit 106 stores a spatial operation filter, for example, an isotropic differential operation filter, Laplacian, or the like. Hereinafter, an example using ΔG described above will be described, but DOG may be used depending on the mode of use.

The PC 100 has the following hardware configuration.
A CPU 201 (see FIG. 2) for performing operations for calculating an evaluation value of image quality including execution of a convolution operation, and a RAM 20 which is a work area of the CPU 201
2 and a hard disk 203 that stores various programs or software including an OS, and stores image data of an image to be evaluated, a spatial operation filter for performing a convolution operation, and a coefficient for an operation.

The PC 100 includes a CRT 204 for displaying an image to be evaluated, a video card 205 for controlling the output of the CRT 204, a printer 206 for printing the image to be evaluated, and a keyboard 2 for giving various instructions.
I / F that controls input and output of a printer 07, a mouse 208, and a printer 206, a keyboard 207, and a mouse 208.
209 and a scanner 210 for reading an image to be evaluated
And a bus 211 for connecting the above components.

Note that the CRT 204 and the printer 206
A predetermined image to be evaluated is displayed or output, which may be used when determining various coefficients or for sampling for performing regression analysis. Similarly, the scanner 210 may be used for sampling. In other words, the input device and the output device can be evaluated by the PC 100.

Here, the storage target of the hard disk 203 will be described. The hard disk 203 is a PC 10
It has an OS 231 for controlling the basic operation of 0, and an image quality evaluation application 232 which is a program for performing a convolution operation, reading various filters and coefficients, and outputting an evaluation value of image quality.

The hard disk 203 includes an image data storage unit 233 for storing image data to be evaluated, a filter storage unit 234 to be referred to when performing a convolution operation,
Coefficient storage unit 2 for storing coefficients used when performing a convolution operation and coefficients used when evaluating the quality of an image
35. As shown in the figure, the image data storage unit 233 stores evaluation target image data 1, evaluation target image data 2,..., And the filter storage unit 234 stores
The differential operation filter 1, the differential operation filter 2,... Are stored, and the coefficient storage unit 235 stores the coefficient 1, the coefficient 2,.
・ Is stored. It is to be noted that the coefficient, which will be described later t ₁ ~
t _5, and stores the _{A 1 ~A 5, p 1 ~p} 9, q 1 ~q 4 or the like.

Next, the contents of each section will be described. (Image Evaluation Device 100: Contents of Image Data Input Unit 101) The image data input unit 101 inputs image data of an image to be evaluated. The image data input unit 101 inputs an image having an image input density approximately equal to the image formation density on the human retina when a human observes the evaluated image at a predetermined observation distance such as 300 mm. Note that an image created electronically in advance may be input. The function of the image data input unit 101 can be realized by, for example, the scanner 210 or the hard disk 203 (in particular, the image data storage unit 233). In addition, depending on the mode of use, it can also be realized by a microdensitometer (not shown).

(Image Evaluation Apparatus 100: Contents of Arithmetic Filter Selection Unit 102) The arithmetic filter selection unit 102 selects an arithmetic filter to be used for the convolution operation.
Hand over to 03. The operation filter is G or Δ
G (see Equation (3) or Equation (4)) is used. In the expression G or ΔG, σ is a constant determined according to each channel. In the case of a Mar model in which there are five channels, there are five σ. Hereinafter, they will be referred to as σ _{1 to} σ ₅ (σ ₁ minimum, σ ₅ maximum) as appropriate.

Therefore, the calculation filter selection unit 102 may select all the filters corresponding to σ depending on the image to be evaluated, or select one σ, for example, a filter corresponding to σ ₂ , depending on the image to be evaluated. There is also. The function of the calculation filter selection unit 102 can be realized by, for example, the hard disk 203 (in particular, the image quality evaluation application 232 and the filter storage unit 234) and the CPU 201. The selected operation filter is stored in the operation filter storage unit 106, and the function of the operation filter storage unit 106 can be realized by the hard disk 203 (in particular, the filter storage unit 234).

(Image evaluation device 100: convolution operation unit 103)
The convolution operation unit 103 outputs the basic data used when applying the selected operation filter to the input image data and calculating the evaluation value of the image quality. That is, the convolution operation unit 103 performs a convolution operation on the image data. In an example described later, a physical quantity expressing a texture of an image is obtained as basic data. Convolution operation unit 103
Are, for example, the hard disk 203 (especially the image quality evaluation application 232), the CPU 201, the RAM 2
02 can realize the function.

(Image Evaluation Apparatus 100: Image Quality Evaluation Unit 1)
04) The image quality evaluation unit 104 outputs an image quality evaluation value based on the result of the convolution operation. In an example described later, an evaluation value correlated with the subjective evaluation of the image is output using the texture amount and the dot area ratio. Specifically, it outputs an evaluation value D represented by Expression (1). The coefficients p ₁ to p ₉ used at this time are coefficients determined in advance, and are stored in the coefficient storage unit 235. Image quality evaluation unit 104
Is, for example, the hard disk 203 (especially the image quality evaluation application 232 and the coefficient storage unit 235) and the C
The function can be realized by the PU 201 and the RAM 202.

(Image Evaluation Apparatus 100: Contents of Dot Area Ratio Acquisition Unit 105) The dot area ratio acquisition unit 105 obtains the dot area ratio of the image to be evaluated and appropriately calculates the absolute value or average value of the difference. Calculated and output to the image quality evaluation unit 104. Here, the acquisition may be either a case where the halftone dot area ratio is calculated based on the image data input from the image data input unit 101 or a case where a numerical value is input via the keyboard 207 or numeric keys. Means that.

The difference value and the average value are, for example, A
When the same image is output using a printer of the company B and a printer of the company B and compared, the difference value or the average value of the halftone dot area ratios is used. Note that the dot area ratio acquisition unit 105 may calculate the absolute value or the average value of the difference, or the image quality evaluation unit 104
May be calculated. Halftone dot area acquisition unit 10
5 can be realized by the hard disk 203 (especially the image quality evaluation application 232), the scanner 210, the keyboard 207, and the CPU 201, for example.

The PC 100, which is an image evaluation apparatus, can quantify the perceived size of an image defect such as banding that occurs in a specific spatial frequency band by having the above-described configuration, and can realize a human visual system. It is possible to perform an image quality evaluation based on the above characteristics, that is, to output an evaluation value correlated with the subjective evaluation of the image.

(Experimental Example) Next, an experimental example to which the image evaluation apparatus 100 is applied will be described. Here, the order of creation of an image, calculation of a physical quantity, a subjective evaluation experiment, and a correlation between an evaluation value and a subjective evaluation experiment will be described. Here, for the sake of simplicity, image quality deterioration due to a texture change due to a gradation change in a halftone-processed image is quantified (evaluated).
The experiment will be described.

(Experimental Example: Creation of Image) The experiment was conducted at 300 dpi and 600 dpi as an image having a texture structure.
Two types of binary halftone images with a screen angle of 30 degrees and 150 lines each at dpi were used. 150 lines means that the halftone dot is 1
This is one attribute value of an image indicating that 150 pieces are included in an inch. FIG. 3 shows that the screen angle used in the experiment was 30 degrees and the screen angle was 15 degrees.
0 line binary halftone image (300 dpi and 600 dpi)
It is an enlarged view of i).

The halftone image data used for the physical quantity calculation and the subjective evaluation experiment was created using a computer. Here, rectangular image data having uniform input values were used as image data for calculating physical quantities, and image data having two different input values and image data consisting of two adjacent rectangular portions were used as image data used for subjective evaluation. . FIG. 4 is a diagram showing an example of an image of a rectangular image of 8 bits per pixel and having the same value over the entire image area as image data for calculating a physical quantity.

FIG. 5 is a diagram showing an example of an image composed of two adjacent rectangular portions having different input values as image data used for subjective evaluation. FIG. 6 is an enlarged view of the image shown in FIG. In FIG. 5, from the viewpoint of legibility, the left side of the rectangular portion is 160 and the right side is 128.

Images for subjective evaluation created on these computers were output by Kodak's APPROVAL. In addition, a binarized image was output from APPROVAL as shown in FIG.

(Experimental Example: Calculation of Physical Amount) Based on the multi-channel theory of the visual system proposed by Mar, five types of channels were used to calculate a physical amount as a basis for quantifying image quality deterioration. According to Mar, the diameter of the central part of the receptive field (ie, the area on the retina represented by the collection of cones (rods) on the retina connecting to one ganglion cell) is , 0.022 °, 0.073 °, 0.146
°, 0.276 °, and 0.495 ° (note that the viewing angle expected to form an image on the unit area on the retina is constant regardless of the viewing distance, and can be represented by ° as described above). .

Therefore, the point of contact of the eyeball optical system is 17
mm, and the diameter of the cone is 2 [μm], the expected angles are 1.65 each of the cone diameter.
Times, 5.42 times, 10.84 times, 20.46 times, 36.
Equivalent to 72 times. The mathematical description of the channel according to Mar defines that the distance from the center of the receptive field is r (r = s
qrt (x ² + y ² )), which is given by the above equation (4).

At the boundary of the central part of the receptive field, equation (4)
Is 0 on the left side (see FIG. 17).
The value of σ for each type of channel is 1 cone
In this case, 1.166 cone, 3.832
cone, 7.665 cone, 14.464 con
e, 25.962 cone.

Assuming that the observation distance of the image is 350 mm, the image of an object having a resolution of 600 dpi, that is, a 1/600 inch is 0.0021 mm on the retina, which is almost equal to the diameter of the cone. Therefore, an image of 600 dpi can be used as a retinal image approximately as it is.

When an image is convolved with such a channel, a change in intensity on the scale of each channel is detected. For an image with uniform intensity (density), the output is 0 for all pixels, but if there is a change in intensity in the image, a positive or negative value corresponding to the change is output. As an example of convolution, using the following equations (5) and (6), a physical quantity T representing a texture is adopted, and an evaluation value D described later is calculated.

[0064]

(Equation 5)

Here, the suffix i in the equations (5) and (6) corresponds to each channel, and G _i (r) represents G (r) in the channel i. N indicates the number of pixels to be operated. The coefficient A _i for obtaining the linear sum T is a multi-valued image t _i having a one-dimensional sine wave intensity distribution with a spatial frequency λ.
Is calculated, and the linear sum of t _i for various λ is converted into the contrast sensitivity spatial frequency characteristic (VTF: Visu) of the human visual system.
al Transfer Function). The reason why the regression is performed on the VTF is that the VTF is considered to be a function expressing the most basic characteristic among human visual characteristics.
That, A _i can be said to be optimized coefficients in the spatial frequency characteristics of the human visual system.

FIG. 7 is an explanatory diagram showing a function form of VTF, and FIG. 8 is a table showing coefficients A _i determined by performing regression. As is clear from FIG. 8, t _i and T are increasing functions with respect to the variation amount of the intensity of the image. In addition, since regression is performed on the VTF, A _i is a constant corresponding to σ _i on a one-to-one basis, and can be used regardless of the type of image to be evaluated.

(Experimental Example: Subjective Evaluation Experiment) In the subjective evaluation experiment, a sample group in which two rectangular figures having different input values were adjacent was presented to the subject. Subjects classified the sample group into five levels according to the difference in perceived texture. That is, level 1 is "clearly understand the difference", level 2 is "clearly understand the difference", and level 3 is
"Looks a bit different", Level 4 "Looks somewhat different", Level 5 "Looks the same"
As two adjacent images were evaluated. FIG. 9 is a diagram showing an example of output from APPROVAL used in the experiment. As described above, in the subjective evaluation experiment, 30
Any two samples of a binary halftone image at 0 dpi, 150 lines, and a screen angle of 30 ° were placed adjacent to each other.

(Experimental Example: Correlation between Evaluation Value and Subjective Evaluation Experiment) The magnitude of the difference in texture perceived for each sample is represented by a value D given by a function represented by equation (1).
Predicted by As described above, ΔT is the difference between the textures given by Expression (6), that is, the texture amount T (texture amount T
_L ) and the texture amount T for the other sample
Which is the difference between (a texture amount _{T R) (T L -T R} ). Also, Tmean is Tmean = (T _L + T _R ) /
2. Note that an image scanned at 600 dpi was used as it was as R (x, y) when obtaining T.

Similarly, ΔM is the difference between the dot area ratios of the samples known in advance, that is, the dot area ratio M (referred to as the area ratio M _L ) of one adjacent sample and the dot ratio of the other sample. a halftone dot area rate M (the area ratio M _R) the difference between the (M _L -M _R). Also, M mean
Mmean = a _{(M L + M R) /} 2.

The inventor of the present application has found that the difference in the amount of texture (Δ
Equation (1) in the form of a linear sum of an amount obtained by correcting T) with the average texture amount (Tmean) and an amount obtained by correcting the difference (ΔM) between the intensity (brightness) of the image with the average intensity (brightness) (Mmean). Devised. Each of the coefficients p _{1 to} p ₉ was determined by performing a regression on the result of the subjective evaluation experiment. FIG.
Is a table showing the values of the coefficients p ₁ ~p ₉ obtained by optimization.

Since the coefficient p3 is negative, the first term of the equation (1) is an increasing function with respect to the average texture amount (Tmean), and corresponds to the phenomenon that the greater the fluctuation of the image intensity, the easier the texture is perceived. It is considered something. The brightness is a decreasing function with respect to the halftone dot area ratio, but since the coefficient p7 is negative, the second term is an increasing function with respect to the average intensity (brightness). It is thought that this corresponds to the phenomenon of being easy. Therefore, it can be said that the evaluation value D calculated by Expression (1) is a prediction expression or evaluation expression that does not contradict the actual psychophysical phenomenon.

(Processing Flow of Image Evaluation Apparatus 100) Finally, the processing flow of the image evaluation apparatus 100 will be described.
FIG. 11 is a flowchart illustrating an example of the processing flow of the image evaluation device 100. In order to evaluate the quality of an image, first, image data of an image to be evaluated is input (step S1101). This input is input to the scanner 210
(See FIG. 2), or ideal electronically created data may be input.

Next, a spatial operation filter to be used for the image data input in step S1101 is selected (step S1102). Examples of the spatial operation filter include an isotropic differential operation filter such as the Mar model shown in Expression (4), and σ _i (i = channel
11 to 5) may be selected, or a spatial operation filter to be used may be selected from a plurality of spatial operation filters different from Expression (4).

Next, a convolution operation is performed on the image data input in step S1101 using the spatial operation filter selected in step S1102 (step S1103). As a specific example, performing an operation represented by Expressions (5) and (6) can be cited. Subsequently, the quality of the image having the image data input in step S1101 is evaluated based on the result of the convolution operation performed in step S1103 (step S1104). As a specific example, obtaining the evaluation value D based on the evaluation expression represented by Expression (1) is given. That is, the evaluation value D is calculated based on the texture amount obtained in step S1103 and the dot area ratio separately acquired.
Ask for.

As described above, the image evaluation apparatus according to the first embodiment can perform image quality evaluation based on the characteristics of the human visual system, and can thereby obtain an evaluation value correlated with the subjective evaluation of an image. An image evaluation device capable of outputting can be provided. In particular, an isotropic differential operation filter ΔG based on a multi-channel model is used as a spatial operation filter, and output values t _i from each channel are used.
Factor A _i for even, so values obtained by regression to VTF, it is possible to obtain an evaluation value that reflects the human visual characteristics.

Embodiment 2 In the second embodiment, the evaluation value D is corrected to further improve the accuracy of the evaluation.
An image evaluation device that outputs an evaluation value highly accurately correlated with human subjective evaluation will be described. In the second embodiment, the same components as those in the first embodiment are denoted by the same reference numerals, and description thereof will be omitted.

FIG. 12 is a functional block diagram when the image evaluation device according to the second embodiment is applied to a PC. The image evaluation device 1200 is an image evaluation device 100 according to the first embodiment.
Further, an image evaluation value correction unit 1201 is included. Next, the contents of the image evaluation value correction unit 1201 will be described.

(Image Evaluation Device 1200: Contents of Image Evaluation Value Correction Unit 1201) The image evaluation value correction unit 1201 converts the evaluation value D output from the image quality evaluation unit 104 into an image evaluation value PD based on the equation (2). To be corrected. The inventor of the present application adopted the function form of Expression (2) to perform logistic regression, and determined coefficients q _{1 to} q ₄ by performing logistic regression. Figure 13 is a table showing the value of the coefficient q ₁ to q ₄ obtained by logistic regression.

FIG. 14 is an explanatory diagram showing the prediction accuracy of the subjective evaluation result by equation (2), that is, the prediction accuracy of the corrected evaluation value. As shown in the figure, the prediction accuracy is a correlation coefficient of 0.901 (contribution ratio of 0.811), and the image evaluation device 1200 outputs an evaluation value that is highly correlated with the subjective evaluation of the image. Turned out to be possible. The image evaluation value correction unit 1201 includes, for example, the hard disk 203 (in particular, the image quality evaluation application 232, the coefficient storage unit 235), the CPU 201, and the RA.
The function can be realized by M202.

(Processing Flow of Image Evaluation Apparatus 1200) Finally, the processing flow of the image evaluation apparatus 100 will be described.
FIG. 15 is a flowchart illustrating an example of the processing flow of the image evaluation device 1200. In order to evaluate the quality of an image, first, image data of an image to be evaluated is input (step S1501). This input is input to the scanner 210
(See FIG. 2), or ideal electronically created data may be input.

Next, a spatial operation filter to be used for the image data input in step S1501 is selected (step S1502). An example of the spatial operation filter is an isotropic differential operation filter as in the Mar model shown in Expression (4), and σ _i (i = channel
11 to 5) may be selected, or a spatial operation filter to be used may be selected from a plurality of spatial operation filters different from Expression (4).

Next, a convolution operation is performed on the image data input in step S1501 using the spatial operation filter selected in step S1502 (step S1503). As a specific example, performing an operation represented by Expressions (5) and (6) can be cited. Subsequently, the quality of the image having the image data input in step S1501 is evaluated based on the result of the convolution operation performed in step S1503 (step S1504). As a specific example, obtaining the evaluation value D based on the evaluation expression represented by Expression (1) is given. That is, the evaluation value D is calculated based on the texture amount obtained in step S1503 and the halftone dot area ratio acquired separately.
Ask for.

The image evaluation apparatus 1200 proceeds to step S15
The evaluation value D calculated in step 04 is corrected to an evaluation value PD based on equation (2) (step S1505). The coefficients q _{1 to} q _{4 for} correction are stored in the coefficient storage unit 235.

As described above, the image evaluation device according to the second embodiment can output an evaluation value highly correlated with the subjective evaluation of an image. For example, in the related art, it is possible to evaluate the detection of an edge using only a texture with good correlation so that a difference is perceived by a human but the difference is not well represented by an evaluation value.

Embodiment 3 Embodiment 3 describes an image evaluation device that adjusts the density of input image data. In the third embodiment, the same components as those in the first embodiment are denoted by the same reference numerals, and description thereof will be omitted.

FIG. 16 is a functional block diagram when the image evaluation apparatus according to the third embodiment is applied to a PC. The image evaluation device 1600 is an image evaluation device 100 according to the first embodiment.
In addition, an input density adjusting unit 1601 is further included. Next, the contents of the input density adjustment unit 1601 will be described.

(Image Evaluation Apparatus 1600: Contents of Input Density Adjusting Unit 1601) The input density adjusting unit 1601 determines the input density of the image data input by the image data input unit 101 by using the retina when a human observes the image. Adjust so as to be approximately equal to the above imaging density. The adjustment is performed by the following method. First, an observation distance when a human observes the evaluated image is input. From the input observation distance, the imaging density on the retina when a human observes the evaluated image is calculated. Next, the image input density is adjusted so as to be approximately equal to the calculated image density on the retina.

For example, assuming that the observation distance when a human observes an image to be evaluated is D [mm], the contact point of the human eye optical system is usually located at a distance of 17 [mm] from the retina and is located on the retina. Since the diameter of the cone is said to be approximately 0.002 [mm], the diameter I of the object imaged on one cone is
[Mm] is represented by I = D · 0.002 / 17. Therefore, the input density adjusting unit 1601 adjusts the image input density to be I [mm] per pixel.

As described above, by adjusting the input density, it is possible to evaluate the image quality based on the characteristics of the actual human visual system. That is, if the observation distance at which the evaluated image is observed by a human is different, the image formation density on the retina is also different. However, the input density adjusting unit 1601 determines that the human can be evaluated at various different observation distances. Can be observed, and an evaluation value having a high correlation can be output. Note that, depending on the mode of the specification, each σ may be adjusted to be enlarged or reduced with the same ratio.

As described above, the present invention uses a model of a visual system channel whose existence is surely recognized from the psychophysical, neurophysiological, and anatomical aspects. As a result, a deductive method based on a real visual mechanism can be introduced into image quality evaluation, and an evaluation value correlated with the subjective evaluation of an image can be output. In other words, according to the present invention, using a channel, which is a mechanism for detecting a change in intensity on the retina, quantifying factors such as graininess, sharpness, and banding that cause image quality deterioration due to image density fluctuation. Can be output as an evaluation value of image quality.

Also, using this evaluation device and evaluation method,
Various image output devices (printer, scanner, CRT monitor) can be evaluated. For example, even if the printers of Company A and Company B use the same 2400 dpi, images printed out from the same image data may have slightly different feelings. In this case, by using the image evaluation device and the image evaluation method of the present invention, it is possible to quantitatively evaluate which is a “good” printer in accordance with a human sensory scale. The reason that such a qualitative variable (human judgment of good or bad image) can be predicted from such a quantitative variable (image data) is that the present apparatus or the present method uses logistic regression analysis. You can say that.

The image processing method described in the present embodiment uses a program prepared in advance for a personal computer.
It can be realized by executing on a computer such as a computer or a workstation. This program is recorded on a computer-readable recording medium such as a hard disk, a floppy (registered trademark) disk, a CD-ROM, an MO, and a DVD, and is executed by being read from the recording medium by the computer. This program can be distributed via the recording medium and a network such as the Internet.

[0093]

As described above, according to the image evaluation apparatus of the present invention (claim 1), the image input means inputs the spatial and optical information of the image, and the convolution operation means inputs the image by the image input means. The spatial image processing unit performs a convolution operation on the spatial and optical information of the obtained image, and the image quality prediction unit predicts the image quality perceived by a human based on the output of the convolution operation unit. It is possible to perform image quality evaluation based on characteristics, and thereby to provide an image evaluation device capable of outputting an evaluation value correlated with a subjective evaluation of an image.

Further, the image evaluation device of the present invention (Claim 2)
In the image evaluation device according to claim 1, since the spatial operation filter is an isotropic differential operation filter, an image quality factor caused by a two-dimensional image structure is quantified based on characteristics of a human visual system. Accordingly, it is possible to provide an image evaluation device capable of outputting an evaluation value correlated with the subjective evaluation of an image.

Further, the image evaluation apparatus of the present invention (Claim 3)
Is the image evaluation device according to claim 1 or 2,
The spatial calculation filter includes a plurality of filters having different characteristics, and the image quality prediction unit uses all outputs from the convolution calculation units corresponding to the plurality of filters, respectively.
Since it predicts the image quality perceived by humans, it is possible to quantify the image quality factors caused by the image structure of various scales based on the characteristics of the human visual system and comprehensively quantify the image quality. An image evaluation device capable of outputting an evaluation value correlated with a subjective evaluation of an image can be provided.

Further, the image evaluation device of the present invention (Claim 4)
In the image evaluation device according to claim 3, since the plurality of filters are optimized for the spatial frequency characteristics of the human visual system, a more precise image quality based on the spatial frequency characteristics of the human visual system Therefore, it is possible to provide an image evaluation apparatus capable of outputting an evaluation value correlated with the subjective evaluation of an image.

Further, the image evaluation device of the present invention (Claim 5)
Is the image evaluation device according to any one of claims 1 to 4, wherein the spatial operation filters have different characteristics.
With the number of filters, it is possible to perform more precise quantification of image quality based on a multi-channel model of the human visual system, thereby outputting an evaluation value correlated with the subjective evaluation of the image. It is possible to provide an image evaluation device capable of performing the following.

Further, the image evaluation apparatus of the present invention (claim 6)
In the image evaluation device according to the third, fourth or fifth aspect, since the filter selecting means selects a required number of the plurality of filters, the image structure is distributed on different visual scales. Perceptions corresponding to multiple types of image defects can be quantified faster and more precisely based on a multi-channel model of the human visual system.
An image evaluation device capable of outputting an evaluation value correlated with a subjective evaluation of an image can be provided.

Further, the image evaluation device of the present invention (claim 7)
In the image evaluation apparatus according to any one of claims 1 to 6, the filter size changing unit changes the size of the spatial operation filter, so that the image quality evaluation independent of the image density of the image to be evaluated can be performed. This makes it possible to provide an image evaluation device capable of outputting an evaluation value correlated with the subjective evaluation of an image.

An image evaluation apparatus according to the present invention (claim 8)
The image data input means inputs image data of an image to be evaluated, and the spatial calculation filter selection means selects a spatial calculation filter to be used for the image data input by the image data input means, and performs convolution. The arithmetic means performs a convolution operation on the image data input by the image data input means using the spatial operation filter selected by the spatial operation filter selection means, and the image quality evaluation means is performed by the convolution operation means. Since the quality of the image is evaluated based on the result of the convolution operation, the image quality can be evaluated based on the characteristics of the human visual system, thereby outputting an evaluation value correlated with the subjective evaluation of the image. It is possible to provide an image evaluation device capable of performing the above.

An image evaluation apparatus according to the present invention (claim 9)
In the image evaluation device according to the eighth aspect, since the spatial operation filter is an isotropic differential operation filter, the image quality factor due to the two-dimensional image structure is quantified based on the characteristics of the human visual system. Accordingly, it is possible to provide an image evaluation device capable of outputting an evaluation value correlated with the subjective evaluation of an image.

The image evaluation device of the present invention (claim 1)
0) is the image evaluation device according to claim 9, wherein the isotropic differential operation filter is set to a function LG (r) = (− 1 / (π)
σ ⁴ )) · (1−r ² / (2σ ² )) · exp (−r ² /
(2σ ² )) (where r is r = sqrt (x ² + y ² )
X and y are coordinates in a coordinate system centered on a certain position of the image data of the image to be evaluated, and σ corresponds to five types of channels in a multi-channel model of a visual system. Is a constant), which allows for more detailed image quality quantification based on a multi-channel model of the human visual system,
An image evaluation device capable of outputting an evaluation value correlated with the subjective evaluation of an image can be provided.

The image evaluation apparatus of the present invention (claim 1)
1) is an image evaluation device that outputs an evaluation according to a subjective evaluation of whether two images look the same or there is a difference by comparing the two images, wherein the image data input means should evaluate The image data of the two images is input, and the texture amount calculation means calculates the absolute value ΔT of the difference between the texture amounts and the average value Tmean of the texture amounts based on the data input by the image data input means. The dot area ratio obtaining means obtains the absolute value ΔM of the difference between the halftone dot area ratios of the two images to be evaluated and the average value Mmean of the halftone dot area ratios. The calculated absolute value ΔT of the difference in the texture amount and the average value Tmean of the texture amount, and the absolute value ΔM of the difference in the halftone dot area ratio and the average value Mmean of the halftone dot area ratio obtained by the halftone dot area ratio obtaining means. And Then, since the image evaluation value D is calculated by the equation (1), it is possible to calculate a difference in appearance due to the texture of an image having the same density macroscopically, thereby obtaining a correlation with the subjective evaluation of the image. And an image evaluation device capable of outputting the evaluated value.

The image evaluation device of the present invention (claim 1)
2) The image evaluation device according to claim 11, wherein the image evaluation value correction unit further converts the image evaluation value D calculated by the image evaluation value calculation unit into an image evaluation value PD based on Expression (2). Therefore, it is possible to provide an image evaluation apparatus capable of outputting an evaluation value highly accurately correlated with the subjective evaluation of an image.

The image evaluation device of the present invention (claim 1)
3) The image evaluation device according to claim 1, wherein the input density adjusting unit determines the input density of the image data input by the image data input unit on the retina when a human observes the image. Since the image density is adjusted so as to be substantially equal to the image density, the image density of the image to be evaluated and the image density on the retina can be made approximately equal, thereby outputting an evaluation value correlated with the subjective evaluation of the image. It is possible to provide an image evaluation device capable of performing the above.

The image evaluation method of the present invention (claim 1)
4) predicts the image quality perceived by a human based on the output obtained by convolving the spatial and optical information of the image with the spatial operation filter, and thus evaluates the image quality based on the characteristics of the human visual system. Accordingly, it is possible to provide an image evaluation method capable of outputting an evaluation value correlated with the subjective evaluation of an image.

The image evaluation method of the present invention (claim 1)
5) In the image evaluation method according to claim 14, the spatial operation filter is an isotropic differential operation filter.
Provided is an image evaluation method capable of quantifying an image quality factor caused by a two-dimensional image structure based on characteristics of a human visual system, thereby outputting an evaluation value correlated with a subjective evaluation of an image. can do.

The image evaluation method of the present invention (claim 1)
6) In the image evaluation method according to claim 14 or 15, all of the outputs obtained by convolving with the spatial operation filter group using a plurality of spatial operation filter groups each having a different characteristic as the spatial operation filter. Using,
Since it predicts the image quality perceived by humans, it is possible to quantify the image quality factors caused by the image structure of various scales based on the characteristics of the human visual system and comprehensively quantify the image quality. An image evaluation method capable of outputting an evaluation value correlated with a subjective evaluation of an image can be provided.

The image evaluation method of the present invention (claim 1)
7) In the image evaluation method according to claim 16, since the spatial operation filter group is optimized for the spatial frequency characteristics of the human visual system, more refinement based on the spatial frequency characteristics of the human visual system. Thus, it is possible to provide an image evaluation method capable of outputting an evaluation value correlated with a subjective evaluation of an image.

The image evaluation method of the present invention (claim 1)
8) In the image evaluation method according to any one of claims 14 to 17, since five spatial operation filters having different characteristics are used as the spatial operation filters, a multi-channel model of a human visual system is used. It is possible to perform a more precise quantification of the image quality based on the image evaluation, thereby providing an image evaluation method capable of outputting an evaluation value correlated with the subjective evaluation of the image.

The image evaluation method of the present invention (claim 1)
In 9), in the image evaluation method according to claim 16, 17 or 18, a necessary number of spatial operation filters are selected and used, so that the image structure is distributed on different visual scales. Perceptions corresponding to multiple types of image defects can be quantified faster and more precisely based on a multi-channel model of the human visual system.
An image evaluation method capable of outputting an evaluation value correlated with a subjective evaluation of an image can be provided.

The image evaluation method of the present invention (claim 2)
0) In the image evaluation method according to any one of claims 14 to 19, since the size of the spatial operation filter is variable, image quality evaluation independent of the image density of the evaluated image can be performed. Accordingly, it is possible to provide an image evaluation method capable of outputting an evaluation value correlated with the subjective evaluation of an image.

Further, the program of the present invention (claim 21)
Is a program that causes a computer to execute image evaluation, and causes the computer to function as each unit according to any one of claims 1 to 13. Each means described in one can be executed by a computer.

[Brief description of the drawings]

FIG. 1 is a functional block diagram when an image evaluation device according to a first embodiment is applied to a PC.

FIG. 2 is an explanatory diagram illustrating an example of a hardware configuration of the image evaluation device according to the first embodiment;

FIG. 3 is an enlarged view of a binary halftone image (300 dpi and 600 dpi) with a screen angle of 30 degrees and 150 lines used in the experiment.

FIG. 4 shows 1 as image data for calculating a physical quantity.
FIG. 7 is a diagram illustrating an example of an image of a rectangular image having 8 bits of pixels and having the same value over the entire image.

FIG. 5 is a diagram illustrating an example of an image including two adjacent rectangular portions having different input values as image data used for subjective evaluation.

FIG. 6 is an enlarged view of the image shown in FIG. 5;

FIG. 7 is an explanatory diagram showing a function form of a VTF.

FIG. 8 is a table showing coefficients A _i used in calculating the texture amount T used in the image evaluation device according to the first embodiment, which coefficients are determined by performing regression.

FIG. 9 is a diagram showing an output example from APPROVAL used in the subjective evaluation experiment described in the first embodiment.

FIG. 10 is a chart showing coefficients p _{1 to} p ₉ which are coefficients used when calculating an evaluation value D used in the image evaluation apparatus according to the first embodiment and which are obtained by optimization; It is.

FIG. 11 is a flowchart illustrating an example of a processing flow of the image evaluation device according to the first embodiment;

FIG. 12 is a functional block diagram when the image evaluation apparatus according to the second embodiment is applied to a PC.

FIG. 13 is a table showing coefficients q _{1 to} q ₄ obtained by logistic regression, which are coefficients used when calculating an evaluation value PD used in the image evaluation apparatus according to the second embodiment.

FIG. 14 is an explanatory diagram showing prediction accuracy of an evaluation value corrected by the image evaluation device according to the second embodiment.

FIG. 15 is a flowchart illustrating an example of a processing flow of the image evaluation device according to the second embodiment;

FIG. 16 is a functional block diagram when the image evaluation device according to the third embodiment is applied to a PC.

FIG. 17 is an explanatory diagram showing a schematic shape of a receptive field of a channel, and is a diagram in which outputs from respective photoreceptors are plotted in a z-axis direction on a receptive field plane and displayed in a horizontal direction.

[Explanation of symbols]

 REFERENCE SIGNS LIST 100 image evaluation device 101 image data input unit 102 operation filter selection unit 103 convolution operation unit 104 image quality evaluation unit 105 halftone dot area ratio acquisition unit 106 operation filter storage unit 203 hard disk 206 printer 210 scanner 232 image quality evaluation application 233 image data Storage unit 234 Filter storage unit 235 Coefficient storage unit 1200 Image evaluation device 1201 Image evaluation value correction unit 1600 Image evaluation device 1601 Input density adjustment unit D Image evaluation value M Halftone dot area ratio PD Evaluation value after correction T Texture amount ΔG Isotropic Sex differential operation filter

Claims (21)

[Claims] 1. An image input unit for inputting spatial and optical information of an image, a spatial operation filter used for the spatial and optical information input by the image input unit, and an input by the image input unit Convolution operation means for performing a convolution operation of the spatial operation filter on the spatial and optical information of the image, and image quality prediction means for predicting image quality perceived by humans based on the output of the convolution operation means. An image evaluation device, comprising: 2. The image evaluation device according to claim 1, wherein the spatial operation filter is an isotropic differential operation filter. 3. The spatial operation filter comprises a plurality of filters having different characteristics, and the image quality prediction means uses all the outputs from the convolution operation means respectively corresponding to the plurality of filters, so that a human can perceive. The image evaluation device according to claim 1, wherein an image quality to be predicted is predicted. 4. The apparatus according to claim 3, wherein the plurality of filters are optimized for spatial frequency characteristics of a human visual system. 5. The image evaluation device according to claim 1, further comprising five filters having different characteristics, respectively, as said spatial operation filter. 6. The image evaluation apparatus according to claim 3, further comprising a filter selection unit that selects a required number of the plurality of filters. 7. The image evaluation apparatus according to claim 1, further comprising: a filter size changing unit that changes a size of the spatial operation filter. 8. An image data input unit for inputting image data of an image to be evaluated, a spatial operation filter selecting unit for selecting a spatial operation filter to be used for the image data input by the image data input unit, Convolution operation means for performing a convolution operation on the image data input by the image data input means using the spatial operation filter selected by the spatial operation filter selection means; convolution operation performed by the convolution operation means An image quality evaluation device, comprising: image quality evaluation means for evaluating the quality of the image based on a result of the calculation. 9. The image evaluation device according to claim 8, wherein the spatial operation filter is an isotropic differential operation filter. 10. The isotropic differential operation filter is defined by a function LG (r) = (− 1 / (πσ ⁴ )) · (1−r ² / (2σ)
² )). Exp (−r ² / (2σ ² )) (where r is r
= Sqrt (x ² + y ² ), where x,
y is a coordinate in a coordinate system centered on a certain position of the image data of the image to be evaluated, and σ is a constant corresponding to five types of channels in a multi-channel model of a visual system.) The image evaluation device according to claim 9, wherein: 11. An image evaluation device for comparing two images and outputting an evaluation in accordance with a subjective evaluation of whether the two images look the same or different from each other, comprising: Image data input means for inputting data; texture amount calculation means for calculating an absolute value ΔT of a difference between texture amounts and an average value Tmean of texture amounts based on the data input by the image data input means; Absolute value Δ of the difference between the dot area ratios of two images to be evaluated
A dot area ratio acquiring unit for acquiring M and an average value Mmean of the dot area ratio; an absolute value ΔT of a difference between the texture amounts calculated by the texture amount calculating unit; and an average value Tmea of the texture amounts
n, the absolute value ΔM of the difference between the halftone dot area ratios obtained by the halftone dot area ratio obtaining means and the average value Mm of the halftone dot area ratios
An image evaluation device comprising: an image evaluation value calculating unit that calculates an image evaluation value D according to the following equation (1) using E.an. (Equation 1) (However, p _{1 to} p ₉ are constants determined by a subjective evaluation experiment performed in advance.) 12. An image evaluation value correction means for correcting the image evaluation value D calculated by the image evaluation value calculation means to an image evaluation value PD based on the following equation (2). The image evaluation device according to claim 11, wherein: (Equation 2) (However, q1 to q4 are constants determined by regression analysis.) 13. An input density adjusting means for adjusting an input density of image data input by said image data input means so as to be substantially equal to an image formation density on a retina when a human observes the image. Claim 1 characterized by the above-mentioned.
13. The image evaluation device according to 12. 14. An image evaluation method, wherein an image quality perceived by a human is predicted based on an output obtained by convolving a spatial operation filter with spatial and optical information of an image. 15. The image evaluation method according to claim 14, wherein the spatial operation filter is an isotropic differential operation filter. 16. A plurality of spatial operation filters having different characteristics are used as spatial operation filters, and image quality perceived by humans is predicted using all outputs obtained by convolution by the spatial operation filters. The image evaluation method according to claim 14, wherein the evaluation is performed. 17. The image evaluation method according to claim 16, wherein the spatial operation filter group is optimized for a spatial frequency characteristic of a human visual system. 18. The image evaluation method according to claim 14, wherein five spatial operation filters having different characteristics are used as the spatial operation filters. 19. The apparatus according to claim 16, wherein a necessary number of the spatial operation filters are selected and used.
19. The image evaluation method according to 7 or 18. 20. The image evaluation method according to claim 14, wherein the size of each of the spatial operation filters is variable. 21. A program for causing a computer to execute image evaluation, wherein the program causes a computer to function as each unit according to any one of claims 1 to 13.