JP4715995B2 - Banding noise detection device, banding noise detection method, and banding noise detection program - Google Patents

Description

  The present invention relates to a banding noise detection device, a banding noise detection method, and a banding noise detection program.

Conventionally, as this type of banding noise detection device, a device that inputs a physical characteristic of a print image, converts the physical characteristic into a spatial frequency, and calculates a banding noise perception amount is known (for example, Patent Document 1, reference.).
According to this configuration, since the visual sensitivity characteristic with respect to the spatial frequency is taken into account when calculating the banding noise amount, it is possible to predict the banding noise amount having a high correlation with the human visual sensitivity characteristic.
JP 2000-4313 A

  Since the print image formed by the ink jet printer is reproduced by a large number of ink dots, banding noise due to banding and granular noise due to dot graininess are mixed in the print image. Therefore, the converted spatial frequency spectrum also includes a component corresponding to granular noise. Thus, there is a problem that even if the banding noise perception amount is predicted based on the spatial frequency spectrum including the component corresponding to the granular noise, it is not possible to obtain an accurate banding noise perception amount.

  The present invention has been made in view of the above problems, and provides a banding noise detection apparatus, a banding noise detection method, and a banding noise detection program capable of accurately detecting banding noise without being affected by granular noise or the like. For the purpose.

Means and actions / effects for solving the problems

  In order to achieve the above object, according to the first aspect of the present invention, granular noise due to ink dots and banding noise are mixed in the printed image, and the image input means inputs the two-dimensional physical characteristics of the printed image. The converting means converts the two-dimensional physical characteristic into a spectrum represented by a two-dimensional spatial frequency. The threshold calculation means smoothes components belonging to a predetermined reference region in the spectrum. Then, the threshold value is calculated by multiplying the smoothed component belonging to the reference region by k times (k is a real number of 1 or more). The banding noise component extraction means compares the spectrum with the threshold, and extracts a component that is equal to or higher than the threshold from the spectrum as a banding noise component.

  Since the threshold is obtained by smoothing the spectrum, a spectrum component that is averagely distributed in the spatial frequency can be represented by the threshold. On the other hand, since the ink dots are uniformly distributed over the entire printed image, the spectral components corresponding to the granular noise are averagely distributed in the spatial frequency. That is, the threshold can be regarded as representing the granular noise at a spatial frequency. Therefore, only the banding noise component can be extracted by comparing the threshold representing the granular noise with the spectrum representing the total noise. Therefore, by analyzing the extracted banding noise component, the banding can be accurately detected without being affected by the granular noise.

  Further, by extracting the banding component by the threshold, the banding component can be accurately extracted even when the direction of occurrence of the banding is unknown. That is, even when the input angle of the print image is shifted, the banding component can be extracted without any problem. The reference area for calculating the threshold can be an arbitrary area in the spatial frequency. Further, the extraction level based on the threshold can be adjusted by multiplying the smoothed reference region component by k (k is a real number of 1 or more). That is, a desired extraction level can be set such that k is increased when the threshold is set high, and k is decreased when the threshold is set low.

  Here, the physical characteristics are not limited as long as granular noise and banding noise can be physically grasped. For example, the brightness, saturation, hue, reflectance, etc. of the printed image can be applied. Further, as the image input means for inputting the physical characteristics in two dimensions, a scanner, a camera, a colorimeter or the like can be used. Further, Fourier transform can be used as the converting means. The spectrum obtained by Fourier transform is not limited to the direct analysis. For example, the power spectrum may be calculated by squaring the spectrum and analyzed. The analysis may be performed after weighting based on various spatial frequency characteristics.

  Furthermore, as an example of a suitable technique for extracting the banding noise component with high accuracy, in the invention according to claim 2, the region extracting means extracts a component included in a predetermined extraction region in the spectrum. Then, the banding noise component extraction means compares the component belonging to the extraction region with the threshold, and extracts a component equal to or higher than the threshold as the banding noise component. That is, since the banding noise is generated in the main scanning direction of the printing apparatus, the banding noise component can be specified in advance in the spatial frequency. Accordingly, since the banding noise component extraction unit can limit the region from which the banding noise component is extracted, only the banding noise component can be extracted with higher accuracy.

  The invention according to claim 3 can evaluate the banding noise suitable for the visual sensitivity characteristic by weighting the spectrum based on the visual sensitivity characteristic.

  Furthermore, as a preferable example of the extraction region, the invention according to claim 4 forms the print image by main-scanning the print head. Then, the extraction area can be extracted based on the main scanning direction of the print head. That is, since the banding noise is generated along the main scanning direction, it can be considered that the banding noise component appears at a spatial frequency orthogonal to the main scanning direction. Further, by setting a region within a predetermined angle range from a direction orthogonal to the main scanning direction as the extraction region, the banding noise component can be extracted even when there is a slight shift in the banding noise direction. Is possible.

  Furthermore, as a preferred example of the reference region, the invention according to claim 5 calculates the threshold based on the spectrum included in the entire region of the spatial frequency by setting the reference region as the entire region of the spatial frequency. can do. That is, it is possible to calculate the threshold that can be compared with the spectrum in the entire spatial frequency region.

  Furthermore, as another example of the reference region, the invention according to claim 6 is characterized in that the reference region is a linear region oriented in the main scanning direction of the print head at a spatial frequency, and thereby the threshold is set to the granular noise. It can be calculated based only on the components. That is, since the banding noise component does not appear in the main scanning direction of the print head at a spatial frequency, the spectral component used for calculating the threshold can be limited to the granular noise component. Therefore, the threshold value suitable for extracting the banding noise component can be calculated. Further, by making the reference region linear, it is only necessary to perform one-dimensional smoothing, so that the calculation can be simplified.

Further, according to the present invention, the banding noise perception amount calculation means calculates the banding noise perception amount based on the banding noise component. By doing so, it is possible to predict the amount of human perception with respect to the banding noise by using the banding noise component accurately extracted.

Furthermore, as a specific aspect of the banding noise perceived amount calculation means, the present invention calculates the banding noise amount by integrating the banding noise component. The banding noise perception amount can be calculated by substituting the calculated banding noise amount into a predetermined conversion formula. That is, by obtaining the sum of the banding noise components of all spatial frequencies, an index of the degree of banding noise can be calculated as the amount of banding noise. The banding noise perception amount can be predicted based on the banding noise amount by a conversion formula that defines the correspondence between the banding noise amount and the banding noise perception amount. For example, the conversion equation can be defined based on the correlation between the banding noise perception amount and the banding noise perception amount obtained by actually measuring the banding noise perception amount in advance by psychological evaluation.

Method for separating the granular noise and banding noise as described above is not necessarily limited to a tangible device is also effective as an invention of Methods. In addition, the above-described banding noise detection device may exist alone, or may be used in a state of being incorporated in a certain device. The idea of the invention includes various aspects. Further, it can be changed as appropriate, such as software or hardware.

When the software of the banding noise detection apparatus is implemented as an embodiment of the idea of the invention, it naturally exists and is used on a recording medium on which such software is recorded. As an example, it has identified invention as bar down loading noise detection program. Of course, the recording medium may be a magnetic recording medium, a magneto-optical recording medium, or any recording medium to be developed in the future.

In addition, the duplication stages such as the primary duplication product and the secondary duplication product are equivalent without any question. Although different from the above medium, the communication line is used as a transmission medium when the communication method is used as a supply method, and the present invention is used .

Here, embodiments of the present invention will be described in the following order.
(1) About print images:
(2) Banding noise perception amount calculation processing:
(3) Banding noise perception amount function:
(4) Regarding other embodiments:
(5) Summary:

(1) About print images:
FIG. 1 shows a state in which a print image which is an evaluation object of the present invention is formed. The print image is a two-dimensional image represented by ink dots on the print paper, and can be formed by moving the nozzle for ejecting ink and the print paper relative to each other. In a general ink jet printer, nozzles capable of ejecting ink are formed on a print head provided on the lower surface of a carriage that accommodates ink cartridges for each color, and the carriage is reciprocated in a certain direction to move in the same direction. A plurality of shifted ink dots can be formed. A clearance of a predetermined distance is formed between the print head and the printing paper, and the ink particles ejected through the clearance fly and ink dots are formed at the landing positions. The carriage operation is referred to as main scanning, and the moving direction is referred to as main scanning direction.

  On the other hand, the printer is provided with a paper feeding mechanism for moving the printing paper, and the printing paper can be fed in a certain direction every time the carriage performs main scanning a predetermined number of times. The operation of the printing paper is referred to as sub-scanning, and the movement direction is referred to as sub-scanning direction. The main scanning direction and the sub-scanning direction are generally orthogonal to each other. By combining these, ink dots can be formed at an arbitrary position on the printing paper. Therefore, it is possible to reproduce a two-dimensional image on the printing paper. Further, the paper feed mechanism does not perform sub-scanning while the print head performs main scanning.

  According to this configuration, the ink dots formed by the same nozzle are arranged in the main scanning direction on the printing paper. Therefore, for example, as shown in FIG. 1, when there is a defective nozzle that causes a deviation in the flying direction of the ink particles of K ink, when a print image is formed with dots of only black (K) ink, the ink formed at the shifted position The dots are continuous, and streak noise (banding noise) along the main scanning direction is formed.

  FIG. 2 schematically shows banding noise. In the drawing, ink dots formed at shifted positions are arranged in the main scanning direction, so that a blank area continuous in the main scanning direction is formed and banding noise is generated. In actual printing, printing using only a single ink is rare, and banding noise rarely continues in the main scanning direction in a strict sense. However, since the shifted ink dots are formed periodically, streaking banding noise is felt when the printed image is viewed as a whole. Such banding noise causes a reduction in image quality of the printed image, and it is necessary to adjust nozzles and the like.

  On the other hand, since the print image is formed of a plurality of ink dots, even if the ink dots are properly arranged, granular noise of the ink dots occurs. For example, granular noise can be grasped as light and dark undulations so that the lightness is low at a part formed by overlapping a plurality of ink dots and the lightness is high at a part not covered with ink dots. However, unlike the banding noise, the granular noise is uniformly distributed over the entire printed image and does not have directionality. As described above, granular noise and banding noise having different properties are mixed in the printed image.

(2) Banding noise perception amount calculation processing:
As described above, banding noise and granular noise are mixed in the printed image, and in order to analyze the banding noise, it is necessary to extract only the banding noise component. By converting physical characteristics obtained from an image into a spectrum represented by a spatial frequency using Fourier transform, the characteristics of each noise can be accurately grasped. Therefore, it is common practice to use Fourier transform for noise analysis of printed images. Also in the present invention, banding noise is analyzed using Fourier transform. The details will be described below.

  FIG. 3 is a flowchart showing a flow of processing for calculating the banding noise perception amount. FIG. 4 conceptually shows a hardware configuration of an apparatus for executing the processing, and FIG. 5 conceptually shows a software configuration for executing the processing. In FIG. 4, the device for calculating the banding noise perception amount includes a computer (PC) 10, a printer 20, and a scanner 30, and the computer 10, the printer 20, and the scanner 30 are connected via a USB I / O 16. It is connected. The PC 10 includes a bus 10a, and a hard disk (HDD) 12, a CPU 13, a ROM 14, a RAM 15, and a USB I / O 16 are connected to the bus 10a. The CPU 13 executes calculations according to programs stored in the ROM 14 and the HDD 12 while using the RAM 15 as a work area.

  FIG. 5 conceptually shows each program module realized according to the above program, and each module is in charge of processing necessary for calculating the banding noise perception amount. In general, the printer 20 includes a printing unit 11b for outputting a test print image and an analysis unit 11a for analyzing image data input from the scanner 30. The analysis unit 11a includes an image input unit 11a1, a two-dimensional Fourier transform unit 11a2, a VTF weighting unit 11a3, an extraction unit 11a4, a banding noise component extraction unit 11a5, a banding noise perception amount calculation unit 11a6, and a threshold calculation unit 11a7. .

  Hereinafter, the processing will be described based on the flowchart of FIG. Note that the printing unit 11b inputs the test image data 12b stored in the HDD 12 in the previous stage of step S100, and the printer 20 outputs the test image on the printing paper based on the test image data 12b. The test image only needs to be able to analyze each noise included in the print image. For example, an image that reproduces a uniform color over the entire surface can be applied as the test image. The printer 20 employs the above-described ink jet method, and banding noise and granular noise are mixed in the test image.

  The test image is placed on a flat head type scanner 30, and the test image is captured as electronic data by the scanner 30. In step S100, the image input unit 11a1 inputs image data regarding the test image. At this time, the image data is input as pixel data in which each pixel is expressed by each gradation value of R (red), G (green), and B (blue). In step S110, pixel data expressed by RGB gradation values is converted into lightness, thereby generating image data in which each pixel is expressed by lightness gradation values. Note that when converting the RGB gradation values into the lightness gradation values, the conversion can be performed by a known conversion equation. Through the above processing, desired physical characteristics (lightness) for the test image are obtained. Needless to say, the scanner 30 can generate two-dimensional image data. Therefore, the brightness data generated in step S110 is also two-dimensional image data.

In step S120, the two-dimensional Fourier transform unit 11a2 performs a two-dimensional Fourier transform on the lightness data, thereby calculating a spatial frequency and its spectrum intensity (corresponding to lightness) for the lightness data. Then, calculated by squaring the spectral intensities same calculation, Wiener spectrum WS (f _x, f _y) for the brightness of the test image. Incidentally, f _x represents the spatial frequency in the x direction, f _y represents the spatial frequency in the y direction. The x direction is the main scanning direction when the printer 20 prints the test image, and the y direction is the sub scanning direction at this time. Incidentally, Wiener spectrum WS (f _x, f _y) spectral intensity corresponds to a value obtained by squaring the amplitude of the brightness waves in the test images.

Figure 6 is a Wiener spectrum WS (f _x, f _y) shows. Wiener spectrum WS (f _x, f _y), which can be the f _x axis and f _y axis and spectral intensity axis representing at three-dimensional space orthogonal, f _x-axis and f _y axis in FIG. 6 And viewed from the normal direction of the spatial frequency plane defined by (the axial direction of the spectrum intensity axis). In the figure, what appears specifically on the f _y axis is a spectrum of banding noise components, and what is uniformly distributed in all directions is granular noise. However, since the banding noise component and the granular noise component are mixed, it is not possible to separate them at the present time.

In step S130, the Wiener spectrum WS (f _x, f _y) with respect to the visual sensitivity function VTF (f _x, f _y) by multiplying those that have been squared, the spectrum of the Wiener spectrum WS (f _x, f _y) The intensity is weighted according to the visual sensitivity characteristic. In general, even if it is a lightness wave of the same amplitude existing in a printed image, humans can easily perceive if the frequency is low, but it is difficult to perceive if the frequency is high. In the present embodiment, in order to perform evaluation suitable for such visual sensitivity characteristics, weighting of spectrum intensity according to the visual sensitivity characteristics is performed. Visual sensitivity function VTF (f _x, f _y) as may be applied to functions such as, for example, the following formula (1). In the following formula (1), l (mm) represents the distance between the test image and the observer.

Figure 7 shows the visual sensitivity function VTF (f _x, f _y) characteristics of the graph. In the figure, the horizontal spatial frequency _fr with the origin (f _x , f _y ) = (0, 0) as the center is shown as the horizontal axis. The formula (1) from the visual sensitivity function VTF (f _x, f _y) is seen to have a symmetry in the circumferential direction around the origin. Accordingly, the curved surface shown visual sensitivity function VTF (f _x, f _y) is at the three-dimensional space is consistent with the trajectory when rotating 360 degrees about the vertical axis to the curve of FIG. Then, VTF weighting section 11a3 at step S130 is multiplied by the visual sensitivity function VTF (f _x, f _y) the Wiener spectrum WS (f _x, f _y) to. The following formula (2) shows the calculation formula at this time.
_{AS (f x, f y)} = (VTF (f x, f y)) 2 × WS (f x, f y) ···· (2)

By doing so, it is possible to perform weighting that emphasizes the spectrum of the low frequency region, the correction spectra were fit to the human visual sensitivity characteristic AS (f _x, f _y) can be obtained. Incidentally, the visual sensitivity function VTF (f _x, f _y) is stored in the HDD 12, VTF weighting section 11a3 are used properly reads. Extractor 11a4 step S140 to extract the predetermined extraction area in the correction spectrum _{AS (f x, f y)} . FIG. 6 shows an extraction area to be extracted at this time. In the present embodiment, the extraction region is an inner region surrounded by two straight lines that pass through the origin and are inclined by a predetermined angle (± α / 2) with respect to the f _y axis.

Here, the f _y- axis direction is the sub-scanning direction and is a direction orthogonal to the banding noise. Therefore, in principle, the spectral component of banding noise appears on the f _y axis. However, the spectral component of banding noise does not always appear on the _fy axis due to printing paper misalignment during printing, placement misalignment with respect to the scanner 30, and the like. Therefore, in the present embodiment, by providing an extraction region of ± α / 2 around the _fy axis, it is ensured that a spectral component regarding banding noise is included in the extraction region. Therefore, it is desirable to set the extraction angle α to an angle having a margin in consideration of the paper feeding accuracy during printing, the placement accuracy in the scanner 30, and the like.

Performing threshold calculator 11a7 is corrected spectrum AS (f _x, f _y) the entire spatial frequency domain (f _x, f _y) of the smoothing processing for at step S150. In this smoothing process, various methods can be adopted, but in the present embodiment, the smoothing process is performed by the following equation (3).

FIG. 8 conceptually shows the state of the smoothing process according to the above equation (3). In the above formula _{(3) FS (f x,} f y) represents the spectrum after the smoothing process. In FIG. 8, a smoothing mask of (q + r + 1) rows × (o + p + 1) columns is formed around the object spatial frequency (f _x , f _y ), and all the spectral components included in the smoothing mask are summed up. ing. By multiplying the filter function F (t, s) to the sum, smoothed spectrum FS (f _x, f _y) are calculated. The filter function F (t, s) is represented by the reciprocal of the area of the smoothing mask regardless of the values of t and s.

The smoothing mask can be arbitrarily sized by changing q, r, o, and p (where q, r, o, and p are natural numbers). For example, the smoothed spectrum FS (f _x, f _y) If you want to calculate the data corrupted the can, q so as to increase the smoothing mask, r, o, it may be p set large. Conversely, smoothed spectrum FS (f _x, f _y) If you want to calculate as pointed data, q so as to reduce the smoothing mask, r, o, it may be set smaller to p.

Threshold calculating unit 11a7, the smoothed spectrum FS (f _x, f _y) at step S160 by multiplying the constant k, the threshold Th (f _x, f _y) is calculated. Threshold Th (f _x, f _y) can be represented by the following formula (4). Incidentally, smoothing processing in the present embodiment is performed for the entire region of the spatial frequency plane defined by the f _x axis and f _y axis, the entire spatial frequency domain is equivalent to a reference area according to the present invention.
_{Th (f x, f y)} = k × FS (f x, f y) ···· (4)

In the above formula (4), k is a constant and can be a real number of 1 or more. If the threshold Th (f _x, f _y) is obtained as described above, to extract the banding noise component in step S170. Specifically, the extraction is included in the area corrected spectrum AS (f _x, f _y) spectral intensity and the threshold Th (f _x, f _y) each spatial frequency and spectral intensity (f _x, f _y) for each of the compared to the threshold Th (f _x, f _y) becomes more corrected spectrum aS (f _x, f _y) by extracting only to extract the banding noise component included in the extraction region. Note that by extracting only the extraction region in step S140, it is not necessary to compare all the regions, so that the processing can be performed at high speed. Further, the threshold Th (f _x, f _y) as in this embodiment by previously calculated for all the spatial frequency domain, it becomes possible to perform the comparison extraction region is formed in which area. Further, even if the region is not extracted in step S140, the comparison can be made between all the spatial frequency regions.

FIG. 9 shows how the banding noise component is extracted. The figure, the vertical axis corrected spectrum AS (f _x, f _y) and the threshold Th (f _x, f _y) represents the value of the spectral intensity, and the horizontal axis f _y axis. Since the threshold Th (f _x, f _y) is smoothed, it can be represented by a smooth curve. Corrected spectrum AS (f _x, f _y), as shown in FIG includes a spectrum projected to have become only ingredients the protruding threshold Th (f _x, f _y) is greater than. The threshold Th (f _x, f _y) to extract only the larger component than.

Since the threshold Th (f _x, f _y) is calculated by smoothing the reference region, it means a variation in the average intensity of the spectrum included in the reference area. That is, the threshold Th (f _x, f _y) is represented the spectral intensity for relief of the lightness are averagely distributed in the test image data. It can be said that the undulations of the brightness dispersed on average in the test image data correspond to the granular noise of the ink dots as shown in FIG. That is, the threshold Th (f _x, f _y) represents the spectral intensity of the granular noise component.

Thus, corrected spectrum AS (f _x, f _y) from threshold Th (f _x, f _y) by extracting components above the spectral intensity, so that the granular noise component is simultaneously removed. That is, the protruding spectral component corresponds to a banding noise component in the printed image, and only the banding noise component can be extracted. Accordingly, in the subsequent processing, only the banding noise component can be analyzed, so that the banding noise can be analyzed more accurately.

In FIG. 9, the banding noise component appears on the _fy axis. However, since the input angle of the test image may be shifted as described above, the banding noise component does not always appear on the _fy axis. However, since the extraction region is set with a margin, banding noise components can be extracted in any direction in the extraction region.

Further, when the value of the constant k in the above equation (4) is increased, the value of the threshold Th (f _x , f _y ) is shifted so as to increase as a whole as shown by the broken line in FIG. Therefore, it is possible to extract only a sharp component by increasing the value of the constant k, and conversely, it is possible to extract a more minute protruding component by decreasing the value of the constant k. That is, an appropriate extraction sensitivity can be set by adjusting the constant k. Incidentally, the constant k is set in advance may be set to a predetermined value, the smoothed spectrum FS (f _x, f _y) and the correction spectrum AS (f _x, f _y) is compared with the appropriate value You may make it do.

Next, in step S180, the banding noise perceptual amount calculation unit 11a6 integrates the spectrum intensity of the banding noise component with respect to the spatial frequency (f _x , f _y ). Of course, since the spectral components other than the extraction region are removed in step S140, the spectral components other than the extraction region are not integrated. In step S190, the banding noise perception amount is calculated by substituting the integrated value (referred to as banding noise amount P. The same applies hereinafter) calculated in step S180 into the banding noise perception amount function 12a. Note that the banding noise perception amount function 12a is stored in the HDD 12 in advance, and is read and used as appropriate by the banding noise perception amount calculation unit 11a6. An example of the banding noise perception amount function is shown in the following formula (5).
S = a × P ^b + c (5)

  In the above equation (5), S represents the banding noise perception amount, and a, b, and c represent preset constants. Thus, by calculating the banding noise perception amount S using the banding noise perception amount function, the banding noise perception amount P can be converted into the banding noise perception amount S suitable for human perception. The banding noise amount P is an index for the banding noise, but is merely a value obtained by integrating the amplitude of lightness in the banding noise. Therefore, the banding noise amount P does not necessarily represent the level of banding noise felt by humans. However, since there is some correlation between the banding noise amount P and the level of banding noise perceived by humans, in the present embodiment, a banding noise perception amount function is set as a correlation function of the two by paying attention thereto. Hereinafter, the banding noise perception amount function will be described.

(3) Banding noise perception amount function In calculating the banding noise perception amount function, first, a large number (n) of various test images are prepared, and Steps S100 to S180 are performed for each of these. That is, the banding noise amount P for various test images is calculated. On the other hand, by performing psychological evaluation on the test image for which the banding noise amount P is calculated, the degree of banding noise perceived by humans is quantified as the banding noise perception amount S. As a psychological experiment method at this time, a scoring method, a paired comparison method, a category method, or the like can be adopted. The banding noise perceived amount S can be defined to be a large value according to the degree to which the banding noise is felt severely, for example, with 0 indicating that the banding noise is not noticed at all. Next, the correlation between the banding noise amount P and the banding noise perception amount S obtained for each test image is investigated.

FIG. 10 is a graph showing the relationship between the banding noise amount P and the banding noise perception amount S for many test images. In the figure, the vertical axis indicates the logarithm of the banding noise amount S, and the horizontal axis indicates the logarithm of the banding noise amount P. This shows that the logarithm of the banding noise perception amount S and the logarithm of the banding noise amount P have a linear relationship. Note that by extracting the banding noise component in step S170, the correlation coefficient between the banding noise amount P and the banding noise perception amount S can be increased. Since such a correlation is seen, the relationship between the two can be assumed as in the following formula (6).

In the above equation (6), the subscripts attached to the banding noise perception amount S and the banding noise amount P indicate the identification number of the test image. Further, the above equation (6) shows the relationship between the banding noise perception amount S and the banding noise amount P for n test images, and is equivalent to the above equation (5). Of course, the constants a, b, and c are common to the above formula (5) and the above formula (6). Taking the logarithm of both sides of the above equation (6), it can be arranged as the following equation (7).

Furthermore, the constants a and b can be expressed by the constant c by solving the lowermost stage of the equation (7). Therefore, the above equation (6) can be expressed only by the constant c. Next, the banding noise amount P for each test image is substituted into the above equation (6) expressed only by the constant c. Banding noise perception value S calculated by this as the temporary banding noise perception values S _c. The value of the temporary banding noise perception value S _c varies by a constant c. The calculated actual and banding noise perception values S for each test image obtained by psychological evaluation, a constant c as square error is the smallest of the temporary banding noise perception values S _c for each test image.

  As a result, the constants a, b, and c are calculated, so the relational expression between the banding noise perception amount S and the banding noise amount P assumed as in the above formulas (5) and (6) is determined. To do. Then, the determined equation (5) is stored in the HDD 12 as the banding noise perception amount function 12a. Therefore, thereafter, the banding noise perception amount calculator 11a6 can appropriately read the banding noise perception amount function 12a and calculate the banding noise perception amount S from the banding noise amount P.

  Thus, by calculating the relational expression between the banding noise perception amount S and the banding noise amount P obtained experimentally, the banding noise amount P obtained in steps S100 to S180 is calculated. The banding noise perception amount S can be predicted. In this embodiment, the logarithm of the banding noise perception amount S and the logarithm of the banding noise perception amount P have a linear relationship. However, the definition of the banding noise perception amount S is various, and this is not necessarily the case. It does not become such a relationship. Therefore, the correlation between the banding noise amount P and the banding noise perception amount S may be expressed by another function. In any case, by statistically grasping the correlation between the banding noise amount P and the banding noise perception amount S, the banding noise perception amount S can be accurately calculated.

(4) Other embodiments:
By the way, it can be considered that the granular noise mixed in the test image has no directionality. Thus, the threshold Th (f _x, f _y) representing the spectral intensity of the granular noise component can be considered to have a symmetry in the spatial frequency. Therefore, it is possible to reduce the calculation burden of the smoothing process performed in step S150 using this objectivity. As a specific example, smoothing processing may be performed by the following equation (8) instead of the above equation (3) in step S150.

FIG. 11 shows a reference region when smoothing is performed using the above equation (8). In the above equation (8), smoothing is performed one-dimensionally in the main scanning direction of the print head. That is, the reference area to be smoothed in this embodiment can be represented by f _x axis _{(f y = 0, f x} ≧ 0). By doing in this way, since an integration area becomes one-dimensional, a calculation burden can be reduced. Further, since the main scanning direction of the print head is a direction along the banding noise, it can be considered that a spectrum component corresponding to the banding noise does not appear in the same direction. That is, since it is possible to perform smoothing using only granular noise component, it is possible to calculate the threshold Th (f _x) which is more suitable for the extraction of banding noise component.

However, the threshold Th (f _x) calculated by the equation (8) f for only _x represents the spectral intensity on the axis, is as it exists on the extraction of the two-dimensional region corrected spectrum AS (f _x , f _y ) cannot be compared with the spectral intensity. Therefore, in this embodiment by utilizing the symmetry of granular noise components, it is used to expand f _x on the axis of the threshold Th to (f _x) in a two-dimensional extraction region. According to the target property of the granular noise component, it can be considered that the granular noise component is the target in the circumferential direction centering on the origin in the spectrum distribution of FIG.

すなわち、抽出領域の補正スペクトルＡＳ（ｆ_x，ｆ_y）とスレッショルドＴｈ（ｆ_r）とのスペクトル強度を比較することが可能となり、バンディングノイズ成分のみを抽出することが可能となる。上述のとおり粒状ノイズは方向性を有していないため、どの方向に平滑化を行っても粒状ノイズの空間周波数特性を示す一次元スレッショルドＴｈ（ｆ_r）を算出することができる。従って、バンディングノイズ以外にも特異なノイズが混在することが予め分かっている場合には、バンディングノイズおよび同特異なノイズのスペクトル成分が現れない方向に参考領域の方向を設定することも可能である。 Therefore, by replacing the variable of the one-dimensional threshold Th (f _x ) with the radius f _r (the same definition as the above expression (1)) as shown in the following expression (9), the threshold Th (f _x ) is changed over the entire region. (F _x , f _y ) can be extended and applied.

That is, the correction spectrum AS (f _x, f _y) of the extracted area it is possible to compare the spectral intensity and the threshold Th (f _r), it is possible to extract only the banding noise component. As described above, since the granular noise has no directionality, the one-dimensional threshold Th (f _r ) indicating the spatial frequency characteristics of the granular noise can be calculated regardless of the direction of smoothing. Therefore, if it is known in advance that unique noise other than banding noise is mixed, it is possible to set the direction of the reference region in a direction in which banding noise and spectral components of the same noise do not appear. .

(5) Summary:
As described above, in the present invention, the threshold representing the granular noise component is calculated by smoothing the spectrum of all noises. Then, only the banding noise component can be extracted by comparing the spectrum of all noises with the threshold and extracting a component that is equal to or higher than the threshold from the spectrum. Accordingly, since the banding noise perception amount can be calculated using only the banding noise component, the banding noise perception amount can be accurately calculated.

It shows how printing is performed by an inkjet printer. It is an enlarged view of a printed image. It is a flowchart of a banding noise perception amount calculation process. It is a hardware block diagram for calculating a banding noise perception amount. It is a program block diagram for calculating the amount of banding noise perception. It is a spectrum figure of noise. It is a graph which shows a visual sensitivity function. It is explanatory drawing which shows the mode of a smoothing process. It is a spectrum figure of noise. It is a correlation diagram between the amount of banding noise and the amount of banding noise perception. It is a spectrum figure of noise.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Computer, 10a ... Bus, 11a ... Analysis part, 11a1 ... Image input part, 11a2 ... Two-dimensional Fourier transform part, 11a3 ... VTF weighting part, 11a4 ... Extraction part, 11a5 ... Banding noise component extraction part, 11a6 ... Banding noise Perception amount calculation unit, 11a7 ... threshold calculation unit, 11b ... printing unit, 12 ... HDD, 13 ... CPU, 14 ... ROM, 15 ... RAM, 16 ... USB I / O, α ... extraction angle

Claims (9)

An image input means for inputting two-dimensional physical characteristics of a print image in which both granular noise and banding noise due to ink dots are mixed;
Conversion means for converting the two-dimensional physical properties to the spectrum expressed by the two-dimensional spatial frequency,
A threshold calculating means for calculating a threshold by which smoothes the components belonging to a predetermined reference area to calculate the smoothed spectrum in the spectral, k times the smoothed spectrum (k is 1 or more real.) To,
From the spectrum, and banding noise component extracting means for extracting intensity is greater than the threshold, or the same as comprising component as banding noise component,
A banding noise detection device comprising: The banding noise detection device according to claim 1,
Provided with a region extracting means for extracting a component belonging to a predetermined extraction area in the spectrum,
The banding noise component extracting means, the intensity of the components belonging to the extraction area, the banding noise detection device and extracts the greater than threshold, or the same as comprising component as banding noise component. The banding noise detection device according to claim 1 or 2,
The banding noise detection apparatus, wherein the spectrum is weighted based on a visual sensitivity characteristic. The banding noise detection device according to claim 2 or 3, wherein
The extraction area, the banding noise detecting apparatus characterized by a direction perpendicular to the main scanning direction of the print head when forming the print image by the printing apparatus is a region to be the spatial frequency within a predetermined angular range . It is a banding noise detection apparatus in any one of Claims 1-4,
The banding noise detection device according to claim 1, wherein the reference region is the entire region of the spatial frequency. It is a banding noise detection apparatus in any one of Claims 1-4,
The reference region is, the banding noise detecting device, characterized in that the linear region oriented in a main scanning direction of the print head in the spatial frequency. The banding noise detection device according to any one of claims 1 to 6,
Comprising a banding noise perception value calculating means for calculating the banding noise perception value based on the banding noise component,
The banding noise perception amount calculating means calculates the banding noise amount by integrating the banding noise component with respect to the spatial frequency,
A banding noise detection apparatus that calculates the banding noise perception amount by substituting the banding noise amount into a predetermined conversion formula . An image input process for inputting two-dimensional physical characteristics of a print image in which grain noise due to ink dots and banding noise are mixed,
A conversion step of converting the two-dimensional physical properties to the spectrum expressed by the two-dimensional spatial frequency,
A threshold calculating step of calculating a threshold by said smoothed components belonging to a predetermined reference area in the spectrum to calculate the smoothed spectrum, the smoothed spectrum k times (k is 1 or more real.) To,
From the spectrum, a banding noise component extraction step of extracting a component whose intensity is greater than or equal to the threshold as a banding noise component ;
A banding noise detection method comprising: On the computer,
An image input function for inputting the two-dimensional physical characteristics of a print image in which grain noise and banding noise due to ink dots are mixed,
A conversion function for converting the two-dimensional physical properties to the spectrum expressed by the two-dimensional spatial frequency,
A threshold calculating function for calculating a threshold by said smoothed components belonging to a predetermined reference area in the spectrum to calculate the smoothed spectrum, the smoothed spectrum k times (k is 1 or more real.) To,
From the spectrum , a banding noise component extraction function for extracting a component whose intensity is greater than or equal to the threshold as a banding noise component ;
A banding noise detection program characterized by realizing

Images