JPH07220083A - Digital image evaluation device - Google Patents

Abstract

PURPOSE:To evaluate an image noise even on a digital color image provided with a pattern of screen which generates any kind of halftone by separating the two-dimensional power spectrum of a noise to regular information and irregular information, and correcting them by the spatial frequency characteristic of a visual system for them independently respectively. CONSTITUTION:A two-dimensional spectrum is found by applying two-dimensional orthogonal transformation to chromaticity information or color difference information to sample the feature of image structure by a two-dimensional spectrum arithmetic means 4. Thence, the two-dimensional spectrum is separated to the regular information and the irregular information by a two-dimensional spectrum separation means 5, and moreover, a one-dimensioned spectrum can be obtained by a two-dimensional spectrum one-dimensionalizing arithmetic means 6. The spectrum is multiplied by a function representing the spatial frequency characteristic of a human visual system for the regular and irregular information, respectively. The image evaluation value of the chromaticity or color difference information is calculated from two values obtained by integration, respectively by an image evaluation value arithmetic means 8.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image evaluation apparatus such as a color hard copy.

[0002]

2. Description of the Related Art Image quality evaluation methods include a psychological evaluation for quantifying the degree of visual perception and a physical evaluation for objectively measuring the properties of an image structure. Image noise is an important factor in image quality. RMS using the standard deviation of change in production as a physical representation of image noise
Examples thereof include a Wiener spectrum obtained by Fourier transforming a change in granularity and density. There is also an evaluation method that combines psychological evaluation and physical evaluation.
rea) image, refer to the document “Noise Perc”.
option in Electrophotogra
phy ”, JournaI of Applied P
photographic engineering, V
ol. 5, p. 190-196 (1979), Roge
r P. Dooley and Rodney Sha
The psychological graininess described in w.
There is a Shaw & Dooley algorithm for predicting s) from the measured value of the Wiener spectrum and the average concentration.
The Wiener spectrum WS (f) is the density variation ΔD from the average density obtained by scanning the image with a microdensitometer.
(X) is a set average of square values of a Fourier spectrum obtained by Fourier transform, and the following equation is used in the Shaw & Dooley algorithm.

[0003]

[Equation 1] ···························································· (1.1) where <> is the collective average, x is the position of the image, Δx is the data sampling interval, u is the spatial frequency, p and q are the width and length of the density system slit N represents the number of data and j represents √-1. In addition, in this Shaw & Dooley algorithm, Wiener spectrum WS (f), average density D
_The following equation using _ave gives the psychological granularity (grainin
ess) G is predicted.

[0004]

[Equation 2] (1.2) Here, Δf is the fundamental spatial frequency, and VTF is the spatial frequency characteristic of the visual system.

However, since the above-mentioned algorithm calculates the evaluation value based on the one-dimensional image information, it cannot be applied to a digital image showing halftone by area gradation.
Therefore, the two-dimensional image inertia information is subjected to two-dimensional Fourier transform to obtain a two-dimensional Fourier spectrum, the two-dimensional spectrum is made into one-dimensional, and the VTF of the human visual system is multiplied and integrated to obtain the influence of the synchronization pattern of the digital image. An algorithm excluding the above is described in Japanese Patent Application Laid-Open No. 1-286084. However, even if this algorithm is used, there is a problem that the evaluation value is significantly different from the evaluation value when visually evaluating an image having different screen patterns for generating halftones.

Further, in the case of the above-mentioned method, since it is viewed only by the variation of the density or the lightness, it can be applied only to the lightness information of the monochrome image or the color image information. Therefore, also in this point, there is a problem that the visually evaluated value and the evaluated value are significantly different. In order to improve this point, an image to be evaluated is converted into two-dimensional chromaticity information and the above-mentioned algorithm is applied, as disclosed in JP-A-5-284260. However, the aforementioned discrepancy with the visual evaluation due to the influence of the screen has not been resolved.

[0007]

As described above, although some methods have been devised as digital image noise evaluation methods in the related art, what kind of halftone is generated in a digital color image having a screen pattern. No algorithm has been devised that can accurately evaluate image noise even for images.

Therefore, in order to solve the above problems, an object of the present invention is to provide an image evaluation apparatus capable of evaluating image noise for a digital color image having a screen pattern which produces any halftone. To provide.

[0009]

As shown in FIG. 1, the structure of the present invention is a two-dimensional image information input means 1 for inputting two-dimensional image information including optical information and two-dimensional position information.
An image information storage means 2 for storing the information of the two-dimensional image inertia information input by the two-dimensional image information input means 1 and the result of arithmetic processing of the two-dimensional image information, and the two-dimensional image information input Preprocessing means 3 for performing arithmetic processing for converting the two-dimensional image information manually input by means 1 into two-dimensional color information, and for performing arithmetic processing for converting the two-dimensional color information into two-dimensional chromaticity information. A two-dimensional spectrum calculation means 4 for generating information indicating a two-dimensional spatial frequency distribution of the two-dimensional chromaticity information by performing a two-dimensional orthogonal transformation on the two-dimensional chromaticity information and calculating a two-dimensional spectrum, Two-dimensional spectrum separation means 5 for separating the two-dimensional spectrum calculated by the two-dimensional spectrum calculation means 4 into regular information and irregular information, and the two-dimensional spectrum separation means 5
The two-dimensional spectral one-dimensional calculation means 6 for performing one-dimensional processing on the regular and irregular two-dimensional spectral information separated by to generate information indicating the one-dimensional spatial frequency distribution, and the two-dimensional Spectral one-dimensional calculation means 6
For the regular and irregular chromaticity information spectra that are one-dimensionalized by, the spatial frequency characteristic is calculated by performing a function and a calculation process that represent the spatial frequency characteristic for the regular and irregular information of the human visual system. Spatial frequency characteristic correcting means 7 for correcting, and image evaluation value calculating means for calculating an image evaluation value from two values obtained by integrating the results for the regular and irregular information output from the frequency characteristic correcting means 7, respectively. The image evaluation device is characterized by having 8.

Further, according to the present invention, in the above image evaluation apparatus, the correction of the spatial frequency characteristic of the visual system by the spatial frequency characteristic correcting means 7 is performed by the two-dimensional spectrum one-dimensional calculating means 6
It may be configured to be performed at the stage of the two-dimensional spectrum before the one-dimensionalization processing by. That is, the spatial frequency characteristic correction means 7 may be provided between the two-dimensional spectrum separation means 5 and the two-dimensional spectrum one-dimensionalization calculation means 6. At that time, the function representing the spatial frequency characteristic of the human visual system with respect to the chromaticity information must reflect the two-dimensional characteristic of the human visual system.

In the image evaluation apparatus of the present invention, the preprocessing means 3 may convert the color information into color difference information, and the color difference information may be processed in the same manner as the chromaticity information.

[0012]

According to the present invention, the color digital image is manually stored by the two-dimensional image information input means 1 while also holding the two-dimensional positional relationship information, and stored in the image information storage means 2. The pre-processing means 3 performs arithmetic processing for conversion into two-dimensional chromaticity information or two-dimensional color difference information. After performing this pre-processing, the two-dimensional spectrum calculation means 4 applies a two-dimensional orthogonal transformation to the two-dimensional chromaticity information or the two-dimensional color difference information in order to extract the feature of the image structure to obtain the two-dimensional spectrum. Ask. Next, the two-dimensional spectrum separating means 5 separates the two-dimensional spectrum into regular information and irregular information, and further, the two-dimensional spectrum one-dimensionalization calculating means 6 separates the regular information and irregular information. A two-dimensional spectrum for information is integrated into a ring (doughnut-shaped region) on the spatial frequency plane of the two-dimensional spectrum to obtain a one-dimensional spectrum.

Next, with respect to the spectrum of the regular and irregular chromaticity information or the color difference information which is one-dimensionalized by the two-dimensional spectrum one-dimensionalization calculating means 6, a human being for the regular and irregular information, respectively. Multiply a function that represents the spatial frequency characteristics of the visual system of. Then, the image evaluation value of the chromaticity information or the color difference information is calculated from the two values obtained by integrating the outputs of the spatial frequency characteristic correction means 7 for the regular and irregular information in the image evaluation value calculation means 8. .

As described above, the present invention obtains a two-dimensional spectrum of an image by two-dimensional orthogonal transformation, separates it into regular and irregular information, and performs one-dimensional processing on the information to evaluate the image. Since the value (physical quantity) is calculated, the image can be evaluated in consideration of the influence of the periodic pattern in the digital color image.

The spatial frequency characteristic correcting means 7 comprises a two-dimensional spectrum calculating means 4 and a two-dimensional spectrum one-dimensional calculating means 6.
If the image evaluation value is provided between the two, the image evaluation value becomes more reliable because the two-dimensional human visual system characteristics are reflected and corrected.

[0016]

FIG. 2 is a block diagram showing the schematic arrangement of a color image quality evaluation apparatus according to an embodiment of the present invention. In the present invention, the image to be evaluated needs to be input as optical information including two-dimensional position information. In this embodiment, X
Color matching functions x (λ), y (λ), z in the YZ color system
Two-dimensional scanning colorimeter 10 with spectral sensitivity equivalent to (λ)
Is used. As the two-dimensional scanning colorimeter, in this embodiment, a color image automatic inspection device shown in Japanese Patent Laid-Open No. 62-299971 is modified to have a spectral sensitivity equivalent to the color matching function. . The output of the two-dimensional scanning colorimeter 10 stores a signal indicating the tristimulus values XYZ at each sampling point in a form in which information on the two-dimensional positional relationship is held. The noise evaluation calculation processing unit 13 uses the image memory 1
A series of operations as shown in FIG. 3 is executed on the evaluated image stored in No. 2 to obtain an image evaluation value representing a physical quantity of image noise. A series of arithmetic processing procedures are stored in the arithmetic processing procedure storage unit 15, and the arithmetic control unit 14
Is the noise evaluation calculation processing unit 1 according to the stored procedure.
Control 3 The calculation processing result by the noise evaluation calculation processing unit 13 is output to the output unit 16 such as a CRT display device.

The flow of a series of processes in the noise evaluation calculation processing unit 13, which is the main part in this embodiment, will be described with reference to FIG.

The L ^* a ^* b ^* conversion two-dimensional image information input by the two-dimensional scanning colorimeter 10 includes tristimulus values X (x, y), Y (x, y), Z (x,
represented by y). Here, x and y are information indicating the position where each tristimulus value information is sampled. In this embodiment, the data sampling interval Δx = 10 μm, Δy
= 10 μm, slit width of the densitometer p = 10 μm, length q = 10 μm, number of data samplings in the x direction N = 10
The number of data samplings in the 24 and y directions was set to M = 1024.

This psycho-physical image inertia information X (x, y),
Y (x, y) and Z (x, y) are converted to a uniform color space in order to bring them closer to the psychological amount of human. A well-known method can be adopted for this conversion, and in this embodiment, CIE1 is used.
L ^* a ^* b in (L ^* , a ^* , b ^* ) space of 976
The conversion to ^* was used. The following formula was used for conversion from XYZ to L ^* a ^* b ^* .

L ^* (x, y) = 116 {Y (x, y) / Yn} ¹ / 3-16: Y (x, y) / Yn> 0.008856 (2) L ^* (x, y) = 903.29 {Y (x, y) / Yn}: Y (x, y) /Yn≦0.008856 (3) a ^* (x, y) ) = 200 [{X (x, y) / Xn} ¹ / 3- {Y (x, y) / Yn} ^1/3 ] b ^* (x, y) = 500 [{Y (x, y) / Yn} ¹ / 3- {Z (x, y) / Zn} ^1/3 ]: X (x, y) / Xn> 0.008856 Y (x, y) / Yn> 0.008856 Z (x, y) ) / Zn> 0.008856 (4) where X (x, y) / Xn, Y (x, y) / Yn, Z
When (x, y) / Zn is 0.008856 or less, the corresponding cube root term in formula (4) is calculated as 7.787 {X (x, y) / Xn} +16/116, 7.787. The calculation is performed by substituting {Y (x, y) / Yn} +16/116, 7.787 {Z (x, y) / Zn} +16/116. Here, Xn, Yn, and Zn represent tristimulus values in the XYZ system on the perfect diffuse reflection surface.

Chromaticity information conversion Chromaticity information E from the image information L ^* a ^* b ^* converted above.
Convert to (x, y). The following formula was used as the conversion formula.

E (x, y) = {L ^* (x, y) ² + a ^* (x, y) ² + b ^* (x, y) ² } ^1/2 ... (5) Further, the following formula is used to convert the color difference information ΔE (xy).

ΔE (x, y) = {ΔL ^* (x, y) ² + Δa ^* (x, y) ² + Δb ^* (x, y) ² } ^1/2 ΔL ^* (x, y) = L ^* ( _{^{x, y) -L * 0 Δa}} * (x, y) = a * (x, y) -a * 0 Δb * (x, y) = b * (x, y) -b * 0 ···· (6) Next, the power spectrum of the obtained chromaticity information E (x, y) was obtained by the following processing.

Discrete Two-Dimensional Fourier Transform In the present embodiment, as a two-dimensional orthogonal transform, a discrete two-dimensional Fourier transform is applied to the chromaticity information E (x, y) or the color difference information ΔE (x, y) processed as described above. The conversion was done. Two-dimensional Fourier transform F for discretely sampled chromaticity information E (x, y) or color difference information ΔE (x, y)
_The calculation was performed using the following formula generally known as _E (u, v).

[0025]

[Equation 3] ...... The (7), when the color difference information Delta] E (x y), when the two-dimensional Fourier transform F [delta] _E (u, v) is expressed by the following equation.

[0026]

[Equation 4] (8) where W ₁ = exp (-j2π / N), W ₂ = exp
(-J2π / M), and E ₀ is the average value of all sampled image information E (x, y).

Two-dimensional power spectrum calculation In the case of chromaticity information, P is the two-dimensional power spectrum of the Fourier transform F _E (u, v) obtained by the equation (7).
_E (u, v) is calculated by the following equation.

P _E (u, v) = | F _E (u, v) | ² (9) Similarly, in the case of color difference information, the Fourier transform FΔ _E ( PΔ _E (u, v), which is the two-dimensional power spectrum of u, v), is calculated by the following equation.

PΔ _E (u, v) = | FΔ _E (u, v) | ² ... (10) These values represent the intensity of the two-dimensional spatial frequency F (u, v).

Two-Dimensional Spectral Rule Information and Irregular Information Separation Processing Next, in order to separate rule information of image noise, that is, an image structure component such as a screen from irregular information, so-called random noise component, two-dimensional Power spectrum P
Separation processing of (u, v) is performed. As a result, the two-dimensional power spectrum P (u, v) has a power
It is separated into a spectrum P ₁ (u, v) and a power spectrum P ₂ (u, v) smaller than the threshold. These separated power spectra P ₁ (u, v), P ₂ (u,
v) represents the rule information (screen) component and the irregular (random noise) warning, respectively. In this embodiment, as the threshold value, the total power spectrum P excluding the DC component is used.
Separation processing was performed using the average value of (u, v). Also,
This separation process is commonly performed for both chromaticity information and color difference information.

Calculation of integral of power spectrum with respect to the same spatial frequency Two-dimensional spectrum regular information, two kinds of power spectra P _1E (U, V), P _2E (U, V) obtained by the irregular information separation processing, or P ₁ Δ
Converting _E (u, v) and P ₂ Δ _E (u, v) into polar coordinates, P _1E (fr, θ), P _2E (fr, θ), or P ₁ Δ
_{After E} (fr, θ) and P ₂ Δ _E (fr, θ), the one-dimensionalized power spectrum P is calculated by the following equation.
_1E (f), P _2E (f), or P ₁ Δ _E (f), P
₂ Calculate Δ _E (f).

P _1E (f) = ΣP _E (fr, iΔθ) · Δθ P _2E (f) = ΣP _E (fr, iΔθ) · Δθ (11) P ₁ Δ _E (f) = ΣPΔ _E (fr, iΔθ) · Δθ P ₂ Δ _E (f) = ΣPΔ _E (fr, iΔθ) · Δθ (12) However, Δθ = 2π / n, which is the basic division in the polar coordinate system. Represents an angle, and fr represents a spatial frequency. Note that P _E (f) obtained by the above equations (11) and (12),
PΔ _E (f) represents a donut-shaped integral centered on the origin of the power spectrum space.

Correction by Visual System Spatial Frequency Characteristic and Calculation of Image Evaluation Value Next, one-dimensional representing the spatial frequency characteristic of the image noise of the regular component and the irregular component obtained by the calculation of the equation (11) or (12). Power spectrum P _1E (f), P
_2E (f), or P ₁ Δ _E (f), P ₂ Δ _E (f)
On the other hand, as shown in the following equation, the spatial frequency characteristic VTF ₁ (f) of the visual system with respect to the regular information and the spatial frequency characteristic VTF ₂ (f) of the visual system with respect to the irregular information are respectively multiplied and integrated, respectively. Q _1E and Q _2E or Q ₁ Δ
By calculating a weighted average based on _E and Q ₂ Δ _E , a metric psychological index serving as an image noise index is calculated.

In the case of chromaticity information,

[Equation 5] (13) However, in the case of α + β = 1 color difference information,

[Equation 6] (14) However, α + β = 1 where α and β in the above equation are actually sensory-evaluated using various digital color image samples, and Q _1E ,
Higher evaluation accuracy can be obtained by calculating Q _2E or Q ₁ Δ _E , Q ₂ Δ _E and deriving an appropriate value using a statistical method. In this embodiment, α = β = 0.5 is set for simplicity.

The spatial frequency characteristics VTF ₁ (f) and VTF ₂ (f) of the visual system for the regular information and irregular information used in this embodiment are shown in FIGS. 4 and 5. The spatial frequency characteristics of these visual systems are experimentally derived by using a simulation image in which noise is experimentally superimposed.

The noise evaluation calculation processing section 13 has been described in detail above. The expressions (2) to (14) used for each calculation may be replaced with other expressions or approximate expressions as long as the same result is obtained. Clearly, it can be used.

Further, since the procedure for calculating the equations (2) to (15) stored in the arithmetic processing procedure storage unit 15 is a design matter which can be arbitrarily constructed by a normal computer program technique, those procedures are not included. Detailed sharpness is omitted. Although the noise evaluation calculation processing unit 13 is realized by computer software technology in the present embodiment, it is needless to say that a part or all of the noise evaluation calculation processing unit 13 can be configured as a hardware by an individual circuit.

An example of the result of actually using the image noise evaluation apparatus of the present embodiment described above will be described.

Image evaluation values were calculated for a large number of samples of digital color images produced by processes such as electrophotography and printing, and sensory evaluation was also performed. The image sample to be measured used digital color solid and halftone images. For sensory evaluation, an evaluation value obtained by quantifying the noise level of each image sample by a category evaluation method was searched. As a result, a very high correlation was obtained between the image evaluation value and the sensory evaluation of this example.

[0040]

As described above, according to the present invention,
The two-dimensional power spectrum of noise obtained by orthogonally transforming the chromaticity information or color difference information of a digital color image is further separated into regular information and irregular information, and visual information for regular information is obtained independently of each other. By correcting the spatial frequency characteristics of the system and the spatial frequency characteristics of the visual system for irregular information, a high correlation with the sensory evaluation value is obtained for digital color images with screen patterns that generate any halftone. An image evaluation value that obtains is obtained, and a highly reliable image quality evaluation can be performed on a digital color image that cannot be realized by the conventional evaluation method.

[Brief description of drawings]

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

FIG. 2 is a block diagram showing a schematic configuration of an image evaluation apparatus according to an embodiment of the present invention.

FIG. 3 shows a configuration of a noise evaluation calculation processing section in FIG.

FIG. 4 shows one-dimensionalized power spectra for regular information and irregular information, respectively, according to an embodiment of the present invention.

FIG. 5 illustrates spatial frequency characteristics of the visual system for regular information and spatial frequency characteristics of the visual system for irregular information used in one embodiment of the present invention.

[Explanation of symbols]

1 ... Two-dimensional image information input means, 2 ... Image information storage means,
3 ... Preprocessing means, 4 ... Two-dimensional spectrum calculation means, 5 ...
Two-dimensional spectrum separation means, 6 ... Two-dimensional spectrum one-dimensional calculation means, 7 ... Spare frequency characteristic correction means, 8 ... Image evaluation value calculation means, 10 ... Two-dimensional scanning colorimeter, 12 ... Image memory, 13 ... Noise evaluation calculation processing unit, 14 ... Calculation control unit, 15 ... Calculation processing procedure storage unit, 16 ... Output unit

Claims (4)

[Claims] 1. A two-dimensional image information input means for inputting an image to be evaluated as two-dimensional image information including optical information and two-dimensional position information, and two-dimensional image input by the two-dimensional image information input means. Image information storage means for storing image information and information obtained as a result of arithmetic processing of the two-dimensional image information, and two-dimensional image information input by the two-dimensional image information input means is converted into two-dimensional color information. And a preprocessing means for performing a calculation process for converting the two-dimensional chromaticity information into two-dimensional chromaticity information, and a two-dimensional orthogonal transform for the two-dimensional chromaticity information. By calculating the two-dimensional spectral frequency distribution of the two-dimensional chromaticity information to generate information indicating a two-dimensional spatial frequency distribution, and the information indicating the two-dimensional spatial frequency distribution as regular information Regulation Two-dimensional spectral separation for separating into regular information, and one-dimensionalized spatial and spatial frequency distribution by subjecting the regular and irregular two-dimensional spectral information separated by the two-dimensional spectral separation means to one-dimensional processing. Two-dimensional spectrum one-dimensional calculation means for generating the information, and for the regular and irregular chromaticity information spectra one-dimensionalized by the two-dimensional spectrum one-dimensional calculation means, respectively, of the human visual system. A spatial frequency characteristic correcting means for correcting the spatial frequency characteristic by performing a calculation process and a function representing the spatial frequency characteristic for the regular and irregular information; and an output of the spatial frequency characteristic correcting means for the regular and irregular information. An image evaluation device, comprising: an image evaluation value calculation means for calculating an image evaluation value from two values obtained by integration. 2. The image evaluation apparatus according to claim 1, wherein arithmetic processing for converting the two-dimensional image information input by the two-dimensional image information input means into two-dimensional color information is performed, and the two-dimensional color information is further processed. An image evaluation apparatus comprising preprocessing means for performing a calculation process for converting the color information into two-dimensional color difference information. 3. A two-dimensional image information input means for inputting an image to be evaluated as two-dimensional image information including optical information and two-dimensional position information, and two-dimensional image input by the two-dimensional image information input means. Image information storage means for storing image information and information resulting from arithmetic processing of the two-dimensional image information, and for converting the two-dimensional image information input by the two-dimensional image information input means into two-dimensional color information And a preprocessing means for performing the calculation processing for converting the two-dimensional color information into the two-dimensional chromaticity information, and the two-dimensional spectrum obtained by performing the two-dimensional orthogonal transformation on the two-dimensional chromaticity information. By calculating the two-dimensional spectral frequency distribution of the two-dimensional chromaticity information to generate information indicating a two-dimensional spatial frequency distribution, and the information indicating the two-dimensional spatial frequency distribution as regular information Two-dimensional spectrum separating means for separating into regular information, and regular and irregular two-dimensional spectral information separated by the two-dimensional spectrum separating means with respect to regular and irregular information of the human visual system Spatial frequency characteristic correcting means for correcting the spatial frequency characteristic by performing arithmetic processing with a function representing the two-dimensional spatial frequency characteristic, and for the output of the spatial frequency characteristic correcting means, primary information is given to each of regular and irregular information. A two-dimensional spectrum one-dimensional calculation means for generating information indicating a one-dimensional spatial frequency distribution by applying a reduction process, and integrating the output of the two-dimensional spectrum one-dimensional calculation means for regular and irregular information And an image evaluation value calculating means for calculating an image evaluation value from the two values. 4. The image evaluation apparatus according to claim 3, wherein arithmetic processing for converting the two-dimensional image information input by the two-dimensional image information input means into two-dimensional color information is performed, and the two-dimensional color information is further processed. An image evaluation apparatus comprising preprocessing means for performing a calculation process for converting the color information into two-dimensional color difference information.