JP2008067101A - Image processor, image processing method, and program for performing the method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processor for reducing nonuniformity to be caused by a specified channel without sacrificing color reproducibility, and correcting the nonuniformity in response to a specified image and an individual device, and to provide an image processing method, and a program for performing the method. <P>SOLUTION: The image processor includes: a correction table generating means for generating a correction table to correct a luminance value; and a reading means for reading sheets with a black and white image printed thereon by color, so as to obtain the images by color. The image processor also includes: an average value acquiring means for acquiring an average value by color, based on the luminance value in each image obtained by the reading means; and a standard deviation acquiring means for acquiring a standard deviation value by color, based on the luminance value in each image obtained by the reading means. The correction table is generated, so as to allow the luminance values in the circumference of the average value of a certain color, which is acquired by the average value acquiring means, to be converged after correction. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an image processing apparatus that executes color conversion processing using a scanner, an image processing method, and a program that executes the image processing method.

  In a scanner-mounted device typified by a color MFP (multifunction printer), unevenness may occur at a specific cycle when a document is read. In particular, when uneven reading occurs in the photograph portion of the document, it interferes with the period of halftone dots of the document itself or the period of the screen used in the printer section, and moire occurs in a copy image. Therefore, in the conventional technique, smoothing is performed using filter processing or the like to reduce unevenness.

  In addition, since the amount of light is often related to unevenness at the time of reading, the conventional technology also uses a technology that detects the amount of light using a device that detects the amount of light and corrects the read value based on the detection result. (See Patent Document 1).

JP 10-118022 A

  However, the above-described conventional filter processing has a problem in that, for example, when the period of unevenness is very large, the size of the filter must be increased, resulting in an increase in cost. Further, when an additional device such as a device for detecting the amount of light is introduced, there is a problem that the cost similarly increases.

  In addition, unevenness may exist only in a specific channel of a specific color. FIG. 1 is a diagram illustrating unevenness that occurs in a specific channel.

  The document data 101 is scanned document data expressed in RGB, and objects 102, 103, and 104 exist on the document data 101. Here, paying attention to the object 102, the object 102 is an image 105 with periodic unevenness. Image data 106, 107, and 108 are obtained by decomposing the image 105 into R, G, and B signal channels. When the distribution of the pixel values of these image data is expressed by standard deviation, it can be seen that the channel data differs for each channel. In the example of FIG. 1, the dispersion value of the B signal is large while the dispersion value of the R signal and the G signal is small. Therefore, the unevenness occurring in the object 102 is considered to be caused by the B signal.

  When filtering processing is performed on such data with R, G, B equal amounts, filtering processing is performed even on data that does not cause unevenness. For this reason, the signal value of data that does not cause unevenness also changes, which has a significant effect on color reproducibility.

  In order to reduce such unevenness, there is a method using a one-dimensional LUT (lookup table). FIG. 8 is a graph showing the relationship between the input and output of a one-dimensional LUT used to reduce unevenness.

  In the one-dimensional LUT 801 in FIG. 8A, the input and the output are in the same relationship. When such an LUT is used, the RGB values of input and output do not change, and the unevenness as shown in FIG. 1 remains as it is. Therefore, an LUT in which the high density portion is crushed is prepared like a one-dimensional LUT 802 in FIG. When such an LUT is used, the variation in the signal in the high density portion becomes the same value, so that unevenness can be suppressed. Since the object 102 in FIG. 1 is data of a high density portion, unevenness can be eliminated by using the one-dimensional LUT 802 in FIG. However, if the unevenness is eliminated by such a method, the gradation property of the high density portion is lost, and the color reproducibility is greatly impaired. Although the range in which the color reproducibility is impaired can be narrowed by using the one-dimensional LUT 802 in FIG. 8B only for a specific channel, for example, even if the high density portion is crushed only in the B signal LUT, Since the gradation of blue is lost, the sacrifice of color reproducibility is still great. Therefore, it has been difficult for the conventional technique to achieve both elimination of unevenness and color reproducibility.

  In addition, unevenness that occurs for a specific channel does not occur for all images, and the degree of unevenness varies from device to device. A method of correcting unevenness is desirable. However, the above-described method has a problem in that correction cannot be performed according to a specific image and according to an individual device.

  The present invention has been made in view of such problems. The object is to reduce the unevenness caused by a specific channel without sacrificing color reproducibility, and to correct the unevenness according to a specific image and an individual device, and an image processing method And providing a program for executing the image processing method.

  In order to achieve such an object, the present invention provides an image processing apparatus having a correction table creating means for creating a correction table for correcting a luminance value, wherein a sheet on which a monochrome image is printed is classified by color. It has a reading means for reading and obtaining images according to colors. Further, the image processing apparatus of the present invention includes an average value obtaining unit that obtains an average value for each color from the luminance values in each image obtained by the reading unit, and a luminance value in each image obtained by the reading unit. It has a standard deviation acquisition means for obtaining a standard deviation value for each color. In addition, when the standard deviation value of a certain color acquired by the standard deviation acquisition unit is larger than the standard deviation value of other colors by a threshold or more, the luminance around the average value of the certain color acquired by the average value acquisition unit A correction table is created so that values converge after correction.

  According to the present invention, an image processing apparatus, an image processing method, and an image processing that reduce unevenness due to a specific channel without sacrificing color reproducibility and correct unevenness according to a specific image and an individual device. It becomes possible to provide a program for executing the method.

  Furthermore, since the present invention does not require a large filter or an additional device, the unevenness can be corrected at a lower cost.

  The embodiment of the present invention includes a multidimensional LUT (lookup table), and performs color conversion using the multidimensional LUT. In one embodiment of the present invention, a multi-dimensional LUT created in advance is adjusted according to an individual device, and unevenness is corrected using the re-created multi-dimensional LUT.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
(First embodiment)
FIG. 2 is a schematic configuration diagram of the color MFP according to the present embodiment. An image reading unit 201 including an auto document feeder irradiates a bundled or single original image with a light source (not shown), forms an original reflection image on a fixed image sensor with a lens, and A raster-like image reading signal is obtained as image information. In a normal copying function, the image reading signal is image-processed into a recording signal in the data processing device 205 and sequentially output to the recording device 203 to form an image on a recording medium. This series of processes is a copy operation. In this specification, this process will be described.

  An operator's instruction to the color MFP is input to an input device 206 which is a key operation unit provided in the color MFP, and a series of operations is controlled by a control unit (not shown) in the data processing device 205. More specifically, the control unit includes a CPU that executes processing operations such as various operations, control, and discrimination, a ROM that stores control programs, and a RAM that temporarily stores data during processing operations and input data. Is provided.

  On the other hand, the display device 204 displays the operation input status and the image data being processed. In addition, the storage device 202 can store image data captured by the image reading unit 201.

  A network I / F 207 is an interface for connecting the color MFP and the network. By using this, image data can be received from a PC or the like, processed by the data processing device 205, and printed by the recording device 203. Further, it is also possible to read a document image by the image reading unit 201 and transmit the image data processed by the data processing device 205 to a PC or the like via the network I / F 207. In the present embodiment, print processing and transmission processing will not be described, but the present embodiment can be applied to the present embodiment in the same manner as copy processing.

  FIG. 3 shows a flow of image processing in the color MFP. The image processing shown in FIG. 3 is performed by the data processing device 205.

  First, the scanner 301 scans document data and reads the data to obtain a device-dependent RGB 302. The scanner 301 corresponds to the image reading unit 201 in FIG. Next, in color conversion 1 in step 303, device-dependent RGB 302 is color-converted to common RGB 305. Here, the common RGB is a device-independent color space, and is a standardized RGB having a reversible relationship with a general device-independent color space such as L * a * b *. In color conversion 1 in step 303, a 3D-LUT (three-dimensional lookup table) 304 is used. FIG. 4 shows an example of the 3D-LUT 304. The 3D-LUT 401 in FIG. 4 is an LUT having a size of 18 × 18 × 18. In the color conversion process of step 303, several candidate input grid points are selected for input device-dependent RGB data, and an interpolation operation is performed to obtain an output value. Here, the output value is the common RGB 305. Further, there is no problem even if any means is used for the interpolation calculation.

  In step 306, the image data converted into the common RGB 305 is stored in the storage device 307. In step 308, image data is read from the storage device 307. Next, in step 309, color conversion 2 is performed using the 3D-LUT 310 to obtain a CMYK (multi-value) 311. FIG. 5 shows an example of the 3D-LUT 310. The 3D-LUT 501 in FIG. 5 has a size of 18 × 18 × 18 as in FIG. Comparing the 3D-LUT 401 in FIG. 4 and the 3D-LUT 501 in FIG. 5, the input side has the same three dimensions but the output side has different dimensions. In the color conversion process at step 309, several candidate input grid points are selected for the input common RGB data, and an interpolation operation is performed to obtain an output value. Here, the output value is CMYK (multivalue) 311. As for the interpolation calculation, there is no problem even if any means is used as in the example of step 303.

  Next, in step 312, gamma correction processing is performed on CMYK (multivalue) 311 to perform tone correction, and in step 313, screen processing is performed to obtain CMYK (binary) 314. This data is sent to the printer 315, and the process ends.

  In the present embodiment, the LUT is corrected by paying particular attention to the 3D-LUT 304 used in the color conversion 1 in step 303.

  FIG. 6 shows a flow of LUT correction processing in the present embodiment. The LUT correction process illustrated in FIG. 6 is performed by the data processing device 205.

First, in step 601, a sample image is read. Here, the sample image is a document image for correcting the LUT.
FIG. 7 shows an example of a sample image. In FIG. 7, patch data 702 to 707 are printed on a sheet 701. The patch data 702 to 707 are composed of various density data from low density to high density. This sample image is an image having the same variation in RGB signals. Here, the type and number of patch data are not limited to those illustrated.

  In step 602, the patch data is cut out. Any method may be used for the patch data extraction process. For example, in the case of a sample image as shown in FIG. 7, patch data 702 to 707 are individually taken out and stored in the RAM. Next, in step 603, a patch that is stored in the RAM and that has not been subjected to correction processing is selected. Subsequent processing is performed on the selected individual patch data.

  Since the patch data is uneven image data, in step 604, the pixel values of the RGB colors are averaged for each of the patch data 702 to 707.

  Next, in step 605, based on the average value of the pixel values acquired in step 604, a target grid point is specified using the 3D-LUT 304, and information on the specified target grid point (target grid point information). Is stored in the RAM. Here, the attention lattice point is a lattice point having the largest contribution rate when performing the interpolation calculation. In addition, the grid points are those whose table values do not change when correcting the 3D-LUT described later. When the average value of pixel values of 8-bit image data is (R, G, B) = (0, 15, 30) at 18 × 18 × 18 lattice points as in the 3D-LUT 401 in FIG. Becomes 255 ÷ 17 = 15. Therefore, if the target lattice point is represented by the input lattice point address of the 3D-LUT 401 in FIG. 4, (R, G, B) = (0, 1, 2). That is, in step 605, a target lattice point is specified for each of the patch data 702 to 707.

  On the other hand, in step 606, channel division is performed on the patch data 702 to 707 selected in step 603. Since RGB is used as an example in the present embodiment, the patch data is divided into an R signal, a G signal, and a B signal. Next, at step 607, a dispersion value is calculated for each channel. Here, since the purpose is to examine the degree of variation of the pixel values of the patch data, a standard deviation or the like can be used as long as it is a calculation formula that can examine the degree of variation. Next, in step 608, it is determined whether or not all channels have been processed, that is, whether or not the dispersion value calculation has been completed for all channels. If not, the process returns to step 607 and is still calculated. Calculate the variance value of the unused channels. After calculating the dispersion values of all the channels, in step 609, the dispersion values of the respective channels are compared, the channel having the largest dispersion value is identified, and the identified channel and the dispersion value are stored in the RAM as channel information 610. To do.

  Next, at step 611, the correction target grid point is extracted using the identified grid point of interest and the channel information 610, and stored in the RAM.

  FIG. 9 shows an example of the flow of extraction processing of correction target grid points. 9 is also performed by the data processing device 205 in the same manner as in FIG.

  First, in step 901, attention grid point information stored in the RAM in step 605 is extracted. Next, in step 902, channel information 610 is extracted. Here, the channel specified in step 609 and the variance value calculated in step 607 are extracted. Next, at step 903, the variance value table 904 is extracted.

  FIG. 10 shows an example 1001 of the variance value table 904. The dispersion level represents the size of the dispersion value, and the larger the dispersion level value, the more the dispersion of signal values, that is, data with greater unevenness. The threshold is a threshold for the calculated variance value, and the threshold increases as the variance level increases. In the variance value table 1001, the threshold values have a relationship of a <b <c. For example, when the calculated variance value is greater than the threshold value a and less than or equal to the threshold value b, the variance level is 2. The grid point range represents the number of grid points to be corrected around the target grid point. The variance value table 904 is not limited to the example of FIG.

  Next, at step 905, the dispersion level is determined. Here, with reference to the dispersion value of the channel information 610 extracted in step 902, the dispersion level is determined based on the threshold value information of the dispersion value table 1001. That is, the variance value included in the channel information 610 is compared with the threshold value of the variance value table 904 to obtain the corresponding variance level. Next, in step 906, based on the dispersion level determined in step 905, a grid point range corresponding to the dispersion level is extracted from the dispersion value table 1001. Finally, in step 907, the correction target grid point is determined based on the target grid point and the extracted grid point range information, stored in the RAM, and the process ends. When the correction target grid point is determined, if a value greater than minus or the number of grid points is calculated, it is excluded from the extraction target. For example, when an 18 × 18 × 18 LUT is targeted, lattice points in the range from (0, 0, 0) to (17, 17, 17) can be corrected.

  As an example, a case where the target lattice point is (R, G, B) = (1, 1, 3), the channel of the channel information is a B signal, and the variance value is larger than b and smaller than c will be described. Referring to the dispersion value table 1001, the dispersion level is 3, and the lattice point range is determined to be 2. Since the channel is a B signal, (1, 1, 1), (1, 1, 2), (1, 1, 4), and (1, 1, 5) are correction target lattice points.

  As described above, after the correction target lattice point is determined in step 611, the correction processing of the correction target lattice point is executed in step 612.

FIG. 11 shows an example of the flow of correction processing of a correction target grid point. 11 is also performed by the data processing device 205 in the same manner as in FIG.
First, in step 1101, the target lattice point A is extracted from the RAM. Next, in step 1102, the correction target grid point B is extracted from the RAM. Next, at step 1103, data corresponding to each channel is extracted. In this embodiment, since a table whose output values are RGB is targeted, each data of the R signal, the G signal, and the B signal is extracted from the 3D-LUT 304. Next, in step 1104, the corrected lattice point C of the correction target lattice point is calculated using the target lattice point A and the correction target lattice point B. Here, the corrected lattice point C can be calculated using the following equation.
C = (A + B) / 2
This equation is intended to bring the value of the correction target lattice point closer to the value of the target lattice point. However, since the distance between the target lattice point and the correction target lattice point is not considered in this expression, the difference between the value of the corrected lattice point and the value of the lattice point outside the correction target becomes large. Therefore, the corrected lattice point C can be calculated by the following equation using a weight (in this case, W) taking into account the distance between the target lattice point and the correction target lattice point.
C = (WA + B) / 2
Here, two calculation formulas for the corrected lattice point C are illustrated, but the calculation formulas are not limited to these.

  In step 1105, the corrected grid point C of the correction target grid point is stored in the 3D-LUT 304. Next, in step 1106, it is determined whether or not all channels have been processed. If not, the process returns to step 1103 to extract new channel data. If all channels are processed, it is determined in step 1107 whether all correction target grid points have been processed. If all the correction target grid points have not been processed, new correction target grid points are extracted in step 1102. If all the correction target grid points have been processed, the process ends.

  FIG. 12 shows an example of the 3D-LUT 304 after executing the correction processing of the correction target lattice points. Here, an example in which the 3D-LUT 304 is corrected as the 3D-LUT 401 in FIG. 4 is shown. Further, the target lattice point is (R, G, B) = (1, 1, 3), and the correction target lattice points are (1, 1, 1), (1, 1, 2), (1, 1, 4). ), (1, 1, 5). The 3D-LUT 1201 is a 3D-LUT after smoothing the correction target lattice points. The output values indicated by italics and underline are the output values after the correction processing, and are closer to the target lattice point (R, G, B) = (1, 1, 3) than the 3D-LUT 401 in FIG. Yes.

  After performing the correction processing of the correction target grid point in step 612 in FIG. 6, the gap between the corrected grid points and the pixels around the corrected grid points is increased. Therefore, a smoothing process is performed at step 613.

  FIG. 13 shows an example of the flow of the smoothing process. First, at step 1301, grid points of the 3D-LUT 304 are extracted. Next, in step 1302, it is determined whether or not the extracted lattice point is the lattice point corrected in step 612. Any means such as comparing the table before and after the correction may be used to determine whether or not the lattice point is corrected. If the extracted lattice point is a corrected lattice point, the subsequent processing is not performed, and it is determined in step 1306 whether all the lattice points have been checked. If the extracted grid point is not a corrected grid point, a neighboring grid point is extracted in step 1303. Next, in step 1304, it is determined whether or not the corrected grid point is in the vicinity. If there is no grid point in the vicinity, the process proceeds to step 1306 without performing the smoothing process. If there are grid points in the vicinity, smoothing processing is performed in step 1305. Here, any smoothing process may be used. After performing the smoothing process, it is determined in step 1306 whether or not all the lattice points have been checked. If not, the process returns to step 1301 to extract the next lattice point. If all grid points are checked, the process is terminated.

  With the above processing, it is possible to reduce a gap between pixel values generated between a corrected lattice point and an uncorrected lattice point. If no gap is generated when the correction target grid point is corrected, the smoothing process may be omitted.

  After performing the smoothing process in this manner, in step 614, the corrected 3D-LUT 304 is stored. Next, in step 615, it is determined whether or not all patch data has been processed. If there is patch data that has not yet been processed, the process returns to step 603 to select a patch that has not been subjected to correction processing. If all patch data has been processed, the process ends.

  Although the copy operation has been described as an example in the present embodiment, any processing such as transmission processing can be applied as long as the processing uses a 3D-LUT that converts device-dependent RGB into common RGB. Further, any LUT can be applied as long as the LUT used for color conversion is multidimensional. For example, the present invention can be applied to a CMYK four-dimensional LUT.

  As described above, according to the present embodiment, by correcting specific grid points of the 3D-LUT, it is possible to reduce unevenness that occurs during scanning without sacrificing color reproducibility. . Further, by processing the channel specified by examining the degree of variation in pixel values, it is possible to reduce unevenness in accordance with a specific image and an individual device.

(Second Embodiment)
In the first embodiment, the example of correcting the 3D-LUT 304 that converts device-dependent RGB into common RGB has been described. However, in this embodiment, the 3D-LUT 310 that converts common RGB into CMYK (multivalue) is corrected. The case will be described.

  Since the flow of the LUT correction process is as shown in FIG. 6 as in the first embodiment, detailed description thereof is omitted. In the present embodiment, the 3D-LUT 304 in FIG. FIG. 14 shows the 3D-LUT 1401 after executing the correction processing of the correction target grid point in step 612. Here, an example in which the 3D-LUT 304 in FIG. 6 is changed to the 3D-LUT 501 in FIG. 5 is shown. Further, the target lattice point is (R, G, B) = (1, 1, 3), and the correction target lattice points are (1, 1, 1), (1, 1, 2), (1, 1, 4). ), (1, 1, 5). The output values indicated in italics and underline are corrected output values, which are close to the target lattice point (R, G, B) = (1, 1, 3) compared to the 3D-LUT 501 in FIG. .

  In the present embodiment, in particular, the amount of K (black) in the UCR process (under color removal process) can be directly corrected, so that unevenness can be corrected more effectively than in the first embodiment. Become. Furthermore, unevenness correction can be performed not only on a scanned image but also on image data obtained via a driver.

(Other embodiments)
Furthermore, the present invention can be applied to a system constituted by a plurality of devices (for example, a computer, an interface device, a reader, a printer, etc.) or an apparatus (multifunction device, printer, facsimile machine, etc.) comprising a single device. It is also possible.

  Another object of the present invention is to read and execute a program of a system or apparatus (or CPU or MPU) from a storage medium storing a program for realizing the procedure of the flowchart shown in the above-described embodiment. Can also be achieved. In this case, the program itself read from the storage medium realizes the functions of the above-described embodiment. Therefore, this program and a storage medium storing the program also constitute one aspect of the present invention.

  As a storage medium for supplying the program, for example, a floppy (registered trademark) disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a magnetic tape, a nonvolatile memory card, a ROM, or the like is used. Can do.

  The functions of the above-described embodiment are realized by executing the program read by the computer. In addition, an OS (operating system) running on the computer based on the instruction of the program performs part or all of the actual processing, and the functions of the above-described embodiments are realized by the processing. It is.

  Furthermore, the program read from the storage medium may be written to a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer. Thereafter, the CPU of the function expansion board or function expansion unit performs part or all of the actual processing based on the instructions of the program, and the functions of the above-described embodiments are realized by the processing.

It is the figure explaining the conventional nonuniformity which generate | occur | produces in a specific channel. 1 is a schematic configuration diagram of a color MFP according to an embodiment of the present invention. FIG. FIG. 6 is a diagram illustrating a flow of image processing in a color MFP according to an embodiment of the present invention. It is the figure which showed the example of LUT used by the process which converts device dependence RGB into common RGB. It is the figure which showed the example of LUT used by the process which converts common RGB into CMYK. It is the figure which showed the flow of the LUT correction process based on one Embodiment of this invention. It is the figure which showed the example of the sample image based on one Embodiment of this invention. It is the figure which showed the example of the one-dimensional LUT used in order to reduce nonuniformity. It is the figure which showed the example of the flow of the extraction process of the correction | amendment object grid point. It is the figure which showed the example of the dispersion | distribution value table. It is the figure which showed the example of the flow of the correction process of the correction | amendment object grid point. It is the figure which showed the example of 3D-LUT after performing the correction process of the correction | amendment object grid point. It is the figure which showed the example of the flow of a smoothing process. It is the figure which showed 3D-LUT after performing the correction process based on one Embodiment of this invention.

Explanation of symbols

304 3D-LUT
610 Channel information (dispersion value / channel)

Claims (5)

An image processing apparatus having a correction table creating means for creating a correction table for correcting a luminance value,
Reading means for reading a printed sheet of a black and white image by color and obtaining an image by color;
Average value obtaining means for obtaining an average value for each color from the luminance values in each image obtained by the reading means;
Standard deviation obtaining means for obtaining a standard deviation value for each color from the luminance value in each image obtained by the reading means,
When the standard deviation value of a certain color acquired by the standard deviation acquisition unit is larger than a threshold value by a standard deviation value of another color acquired by the standard deviation acquisition unit,
The correction table creating means includes
An image processing apparatus that creates a correction table so that luminance values around the average value of the certain color acquired by the average value acquisition unit converge after correction. The correction table creating means includes
When the standard deviation value of a certain color acquired by the standard deviation acquisition unit is larger than a threshold value by a standard deviation value of another color acquired by the standard deviation acquisition unit,
The correction table creating means includes
With the luminance value corresponding to the average value of the other color acquired by the average value acquisition unit as a constraint, the luminance value around the average value of the certain color acquired by the average value acquisition unit is The image processing apparatus according to claim 1, wherein a correction table is created so as to converge after correction. An image processing method having a correction table creation step for creating a correction table for correcting a luminance value,
Reading a printed sheet of a black and white image by color and obtaining an image by color; and
An average value obtaining step of obtaining an average value for each color from the luminance value in each image obtained in the reading step;
A standard deviation obtaining step of obtaining a standard deviation value for each color from the luminance value in each image obtained in the reading step,
When the standard deviation value of a certain color acquired in the standard deviation acquisition step is larger than a threshold value by a standard deviation value of another color acquired in the standard deviation acquisition step,
The correction table creation step includes
An image processing method, wherein a correction table is created so that luminance values around the average value of the certain color acquired in the average value acquisition step converge after correction. The correction table creation step includes
When the standard deviation value of a certain color acquired in the standard deviation acquisition step is larger than a threshold value by a standard deviation value of another color acquired in the standard deviation acquisition step,
The correction table creation step includes
With the luminance value corresponding to the average value of the other color acquired in the average value acquisition step as a constraint, the luminance value around the average value of the certain color acquired in the average value acquisition step is The image processing method according to claim 3, wherein a correction table is created so as to converge after correction.   A program for executing the image processing method according to claim 3.

Images