JP4823973B2 - Image calibration method - Google Patents

Description

  The present invention relates to an image calibration method for determining and calibrating image quality deterioration of an image input / output device such as a scanner, a printer, or a multifunction printer (MFP).

  As a method for evaluating the image quality of a digital image, subjective evaluation using a limit sample or a rank sample is widely used. However, such a method based on subjective judgment has a number of problems such as variations in evaluation results and no guarantee of the accuracy of the sample itself. In order to compensate for these drawbacks, various objective image quality evaluation methods based on physical measurement in which an image is calculated and an evaluation value is calculated have been proposed and put into practical use.

  For example, in Patent Document 1, an original image evaluation value is calculated based on original image data, input image data corresponding to the original image data is taken in to calculate an input image evaluation value, and the calculated original image evaluation value and input By comparing with the image evaluation value, a total evaluation value representing the total image quality of the output image is calculated. It is described that the evaluation value obtained for the image is compared, not the pixel-by-pixel comparison, and the image quality element evaluation value is simply calculated.

  Further, in Patent Document 2, in order to evaluate the quality of the compressed / decompressed image of the original image, each spatial frequency spectrum is obtained, and further, a spatial frequency spectrum of a specific region that affects human visual characteristics is extracted. It is described that this is done by comparison.

  In Patent Document 3, in an image input method in which an image to be evaluated is input and converted into an electrical signal and the input characteristics of the image are corrected, the spatial frequency characteristics of the input image are corrected when the input characteristics of the image are corrected. The correction configuration enables the evaluation of the image quality by inputting the image while suppressing the noise component even in the dark part image while correcting the deterioration in the high spatial frequency region even when using a general-purpose image scanner. Is described.

  In a natural image, the power spectrum is a decreasing function with respect to the spatial frequency, and the power spectrum is substantially proportional to the reciprocal of the spatial frequency when analyzed in more detail. Such characteristics in natural images are considered to be general characteristics (see Non-Patent Document 1).

In order to evaluate the reproducibility of the image structure of the image processing system or image input / output system that causes some change in the image, it is necessary to consider such statistical properties of the natural world. Furthermore, if the reproducibility of the image structure is to be evaluated without depending on the image type, it is desirable to have a spatial structure that is as random as possible.
JP 2006-237676 A JP-A-5-176179 JP 2006-270772 A Fitting the Mind to the World: Adaptation and Aftereffects in High Level Vision: Advances in Visual Cognition Series, Volume 2, C. Clifford and G. Rhodes (Eds.) Oxford University Press.

  When image data is not available due to the spread of personal and business scanners and multifunction printers (MFPs) equipped with scanners, images that have been printed once are read and input by a scanner, and then a hard disk or CD-ROM Opportunities for storing the data in a recording medium such as the above and reusing this data are increasing. However, the image quality is deteriorated by the print output, and further, the image quality is doubled by inputting this image with the scanner. For this reason, when reusing this, there was a problem that an image with very poor image quality would be used.

  The present invention is directed to solving the problems of the prior art, using image data having a spatial frequency characteristic in which a power spectrum that can be regarded as a generalized natural image is approximately proportional to the reciprocal of the spatial frequency, It is an object of the present invention to provide an image quality degradation determination method for determining a factor and a degree of image quality degradation and an image calibration method for improving the image quality in an image input apparatus using the result.

  The image calibration method for an image input device according to claim 1 is an image calibration method for an image input device that reads an image output from the image output device, and the power spectrum is the reciprocal of the spatial frequency together with the main image. A calibration image having a spatial frequency characteristic approximately proportional to is output to the same output medium by the image output device, the output image is input from the image input device, and the input spatial frequency obtained from the input value of the calibration image A one-dimensional power spectrum is obtained by comparing the characteristics with the spatial frequency characteristics of the calibration image, and based on the one-dimensional power spectrum, the image quality deterioration factor of the image input device and the degree of the image quality deterioration are determined, and the determination result is obtained. Based on this, the image calibration procedure at the time of main image input is changed.

  An image calibration method for an image input device according to claims 2 and 3 is the image calibration method according to claim 1, wherein the input spatial frequency characteristic obtained from the input value of the image is compared with the spatial frequency characteristic of the image. Is based on the difference between the input spatial frequency characteristic and the spatial frequency characteristic of the image in the Fourier space by the process of Fourier transform to obtain a one-dimensional power spectrum, and further causes image quality degradation based on the one-dimensional power spectrum. The degree of image quality degradation is determined by comparison with a known interference power spectrum obtained in advance by deconvolution calculation.

  According to the above configuration, an image having a spatial frequency characteristic whose power spectrum, which can be regarded as a generalized natural image, is approximately proportional to the reciprocal of the spatial frequency, and the spatial frequency characteristic data included in the image, are used to cause degradation in image quality. It is possible to improve the image quality of the image input apparatus using the result.

  According to the present invention, by using an image having a spatial frequency characteristic in which a power spectrum obtained by generalizing a natural image is approximately proportional to the reciprocal of the spatial frequency and data of the spatial frequency characteristic, it is possible to determine the deterioration factor of the image quality and its degree. The determination result can be used to improve the image quality of the image input apparatus.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  First, an example of image data used in the present invention will be described with reference to FIGS. FIG. 1 shows one-dimensionalization of the power spectrum of some typical natural images by averaging the power of each point in the spatial frequency for each frequency band. As shown in FIG. 1, in a natural image, the power spectrum is a decreasing function with respect to the spatial frequency, and the power spectrum is substantially proportional to the reciprocal of the spatial frequency when analyzed in more detail.

  As described in the above background art, such a spatial frequency characteristic in a natural image is considered to be a general characteristic. In order to evaluate the reproducibility of the image structure of the image processing system or image input / output system that causes some change in the image, it is necessary to consider such statistical properties of the natural world. In order to evaluate the reproducibility of the image structure without depending on the image type, it is desirable to have a random spatial structure as much as possible.

  In the digital image input / output system, the reproducibility of the image structure, that is, the input image or the image structure such as edges, textures, and shades of all objects constituting the subject in the natural world, after image input, image output or image processing As an image data for evaluating how accurately the image is reproduced, it is an image in which the power spectrum has a spatial frequency characteristic that is approximately proportional to the reciprocal of the spatial frequency. The image can be considered as a generalized image of a natural image.

  In order to generate this image data, a random number generation means, a Fourier transform means, a power spectrum modulation means, and an inverse Fourier transform means are provided. First, the random number generation means generates image data composed of two-dimensional random numbers, and then Fourier transform (FFT) is performed on image data composed of two-dimensional random numbers. The frequency-converted image data is modulated by the power spectrum modulation means so that the power spectrum becomes a decreasing function with respect to the spatial frequency as described above, or so that the power spectrum is substantially proportional to the reciprocal of the spatial frequency. . The result is subjected to inverse Fourier transform (inverse FFT) to generate real space image data (calibration image).

  FIG. 2 is a diagram showing an example of a calibration image having these characteristics. FIG. 3 is a one-dimensional representation of the power spectrum of the calibration image, and the spatial frequency characteristic is approximately proportional to the reciprocal of the spatial frequency.

  Since the sensitivity of the human visual system is limited to about 60 cycles (60 cpd: cycle per degree) at the maximum with respect to an expected angle of 1 degree, the spatial frequency component constituting the generated image data is This level may be at most, and this corresponds to about 11.5 cpm (cycle per mm) when it is assumed that the observation distance is 300 mm, which is a general observation distance (distance for clear vision).

  In a digital image input device using a CCD array or a CMOS array as a light receiving element such as a scanner or a digital camera, there are two main causes of image quality degradation: noise due to variations in sensitivity between light receiving elements and optical blur. It is thought that.

  Inputting an image such as a natural image from such a digital image input device is nothing other than the fact that these noises and blurs are folded into the image as interference. The interference that is convoluted can be explicitly known by dividing by the original image at the spatial frequency, that is, by deconvolution operation.

  However, since the interference that becomes explicit at the spatial frequency differs for each input image, it is difficult to know the average interference that is convoluted by the image input device by a normal method. However, an image having a spatial frequency characteristic in which the power spectrum as shown in FIG. 2 is approximately proportional to the reciprocal of the spatial frequency is considered to be a generalized natural image. Therefore, image data obtained by inputting this image. The interference that is convoluted with respect to can be considered the average interference that is convolved with the natural image by the image input device.

  In addition, if the density or brightness distribution of such an image is known, it is possible to perform a deconvolution calculation. FIG. 4 is obtained by adding the various noises (levels 1 to 8) to the calibration image shown in FIG. 2 whose brightness distribution is known and then dividing the image by the original calibration image at a spatial frequency. That is, it is a graph in which the power spectrum for each noise that is a convoluted disturbance is made one-dimensional.

  FIG. 5 similarly shows the result of adding various blurs (level 1, 4, 8) to the calibration image shown in FIG. It is the graph which made the power spectrum for every certain blur into one dimension. As apparent from FIGS. 4 and 5, the graphs differ greatly depending on the interference that is convoluted, that is, the difference in image quality deterioration factors.

(Embodiment 1)
FIG. 6 is a flowchart showing the processing of the image quality degradation determination method in Embodiment 1 of the present invention. As shown in FIG. 6, first, the image for calibration shown in FIG. 2 having a known brightness distribution (spatial frequency characteristic of the power spectrum) is read by an image input device such as a scanner or a digital camera (S1). Next, the calibration image is read, and the pixel value of the input image is converted into a lightness value by using the input / output characteristics of an image input device that is known in advance (S2). Next, the input image data (lightness distribution) converted to the lightness value is subjected to Fourier transform (FFT) processing (S3). Further, the image data for calibration (lightness distribution) known in advance of the calibration image (FIG. 2) is subjected to Fourier transform processing (S4). Next, the input image data is divided by the calibration image data in the Fourier space (S5). Next, the power of each point in the spatial frequency of the divided image data is averaged for each frequency band, thereby obtaining a one-dimensional power spectrum of the interference convolved by the image input device ( S6).

  By comparing this power spectrum with the known interference power spectrum obtained in advance by the deconvolution calculation shown in FIG. 4 (noise) and FIG. 5 (blur), the power spectrum is convoluted by reading the image input device. It is possible to know the cause and the degree of image degradation due to interference (S7).

  For example, Ni (f) is a known noise power spectrum and Bj (f) is a known blur power spectrum, and ΣPi * Ni (f) + ΣQj * Bj (), which is a linear sum of each known disturbance. f) (where i corresponds to the individual known noise and j corresponds to the individual known blur) to the one-dimensional power spectrum Y (f) of the interference convolved by the image input device. By performing a regression on Pi, it is possible to know Pi and Qi, which are the degree of interference. If one simply wants to know the main image degradation factor and its degree, the correlation coefficient of Y (f) and Ni (f) for all i and Bj (f) for all j are calculated. Choose the one that maximizes.

(Embodiment 2)
FIG. 7 is a flowchart showing a process for changing the image calibration procedure based on the result of the image quality degradation determination method according to the second embodiment of the present invention. Further, in the second embodiment, components having equivalent functions corresponding to the configuration requirements described in the image quality degradation determination method of the first embodiment shown in FIG. 6 are denoted by the same reference numerals.

  As shown in FIG. 7, in the second embodiment, the image calibration procedure at the time of image input is changed based on the result of the image quality deterioration determination similar to that of the first embodiment. When the procedure (processing S1 to S7) shown in the first embodiment reveals the type of disturbance that is averagely convolved with the natural image by the image input device, that is, the image quality deterioration factor, the image is based on the result. The calibration procedure is changed (S8).

  For example, if it is determined in steps S1 to S7 that the main image quality deterioration due to the image input device is blurred, processing such as edge enhancement that corrects the subsequent natural image can be performed to correct the blur. . If it is determined that the main image quality degradation by the image input device is noise, it is possible to perform a process such as smoothing for correcting the natural image that will be input thereafter to make the noise inconspicuous. .

  Next, a normal image input procedure in the image input apparatus after changing the image calibration procedure will be described with reference to FIG. As shown in FIG. 8, an image that the user wants to input is read and input by a digital image input device such as a scanner or a digital camera (S11), and is input by the image calibration procedure changed as described above. The normal image data is corrected for image deterioration (S12). Thereafter, the corrected normal image data is stored in a recording medium such as a hard disk or a CD-ROM and is reused (S13).

  Thus, in the second embodiment, the image calibration procedure is changed based on the revealed image quality deterioration factor. The calibration process using the calibration image of FIG. 2 may be performed every time the image input apparatus is started, or may be performed periodically such as once a month, or image degradation is a concern. In this case, the user may arbitrarily perform it.

(Embodiment 3)
In Embodiment 3 of the present invention, the main image quality deterioration factors in the printer (image output device) are considered to be image modulation, noise, and blur due to halftone processing, so an image input device corresponding to this specific image output device This is a process for changing the image calibration procedure.

  FIG. 9 shows various images of the halftone process (modulation) performed when the data of the calibration image of FIG. 2 having a known brightness distribution in the third embodiment is output on the print medium of the output medium. And is divided by the original calibration image data at the spatial frequency in the same manner as in FIGS. 4 and 5 described above. That is, it is a graph in which the interference power spectrum convoluted by halftone processing is made one-dimensional.

  FIG. 10 is a flowchart showing processing for changing the image calibration procedure of the image input apparatus corresponding to an arbitrary image output apparatus. In FIG. 10, a target image output device (printer) outputs the data on the print medium with the calibration image data having the known brightness distribution shown in FIG. 2 (S20). At this time, as described above, image degradation such as modulation, noise, and blur occurs in the image by the halftone processing.

  Next, the output image is read and input by a digital image input device such as a scanner (S21). The pixel value of the read and input image is converted into a brightness value using the input / output characteristics of an image input device that is known in advance (S22). The input image data (lightness distribution) converted to the lightness value is subjected to Fourier transform processing (S23). In addition, the data of the calibration image having the known brightness distribution shown in FIG. 2 is subjected to Fourier transform processing (S24). Next, the input image data is divided by the calibration image data in Fourier space (S25).

  Next, the power of each point in the spatial frequency of the divided image data is averaged for each frequency band to be one-dimensional, and the one-dimensional power of the interference convoluted by the image output device and the image input device is obtained. A spectrum is obtained (S26).

  By comparing the one-dimensional power spectrum obtained in this way with the known power spectrum obtained in advance as shown in FIG. 4, FIG. 5 and FIG. In addition, it is possible to know the image deterioration factor due to the interference convoluted by the image input device and its degree (S27).

  Based on this result, the image calibration procedure at the time of image input corresponding to the target image output apparatus is changed (S28). For example, when the main image quality degradation by the image output device and the image input device is modulation of an image by halftone processing, if details of the halftone processing (number of screen lines, resolution, etc.) are clarified in the above procedure, Thereafter, for an image output from the same image output device and input to the image input device, for example, halftone processing is performed according to the result of the procedure using the calibration image in FIG. The influence of modulation can be reduced.

  For details of halftone processing by the image output apparatus, for example, the power spectrum of modulation by known halftone processing as shown in FIG. 9 is Hi (f), and convolved with Hi (f) for all i. The correlation coefficient of the one-dimensionalized power spectrum Y (f) of the jammed interference is calculated, and the one that maximizes the correlation coefficient may be selected. As the halftone dot removal processing, smoothing processing may be performed by a method according to the halftone processing estimated here.

  In addition, by performing the processing of the third embodiment after changing the image calibration procedure in the above-described second embodiment, it is possible to more accurately identify the cause of the image deterioration in the halftone processing in the image output apparatus. it can.

  Next, a normal image input procedure corresponding to a specific image output device in the image input device after changing the image calibration procedure will be described with reference to FIG. As shown in FIG. 11, an image is output by the image output device (S30), the output image is input by a digital image input device such as a scanner or a digital camera (S31), and the image calibration changed as described above is performed. Image degradation correction is performed on the image data input by the procedure (S32). Thereafter, the corrected image data is stored in a recording medium such as a hard disk or a CD-ROM, and is reused (S33).

  Thus, in the third embodiment, the image calibration procedure is changed based on the revealed image quality deterioration factor. In particular, when the output from a specific image output device is input by the image input device, the calibration processing using the calibration image data shown in FIG. 2 may be performed every time the image input device or the image output device is started, for example. It may be performed periodically such as once a month, or may be arbitrarily performed by the user when image degradation is a concern.

(Embodiment 4)
FIG. 12 is a diagram showing an image output from the target image output apparatus (printer) and output on the same print medium input by the image input apparatus in Embodiment 4 of the present invention.

  In the second and third embodiments described above, the calibration image data whose brightness distribution is known in FIG. 2 and the normal image data that is actually saved and reused by the image input device are output separately on the print medium. It was read and input separately into the image input device. In the fourth embodiment, as shown in FIG. 12, a calibration image 5a and a normal image 5b are output from the target image output device onto the same print medium 5.

  FIG. 13 is a flowchart showing processing for changing the image calibration procedure of the image input apparatus corresponding to an arbitrary image output apparatus according to the fourth embodiment. In FIG. 13, a calibration image 5a having a known brightness distribution shown in FIG. 2 is displayed on the same print medium 5 from a target image output apparatus (printer) together with a normal image 5b actually used for actual use. (S40). At this time, as described above, image degradation such as modulation, noise, and blur occurs in the image by the halftone processing.

  Next, all the images (5a, 5b) output on the printing medium 5 are read and input by a digital image input device such as a scanner (S41). The pixel value of the read calibration image 5a is converted into a lightness value using the input / output characteristics of an image input device that is known in advance (S42). The input image data (lightness distribution) converted to the lightness value is subjected to Fourier transform processing (S43). Further, the image data for calibration whose brightness distribution shown in FIG. 2 is known in advance is subjected to Fourier transform processing (S44). Next, the input image data is divided by the calibration image data in Fourier space (S45).

  Next, the power of each point in the spatial frequency of the divided image data is averaged for each frequency band to be one-dimensional, and the one-dimensional power of the interference convoluted by the image output device and the image input device is obtained. A spectrum is obtained (S46).

  By comparing the one-dimensional power spectrum obtained in this way with the known power spectrum obtained in advance as shown in FIG. 4, FIG. 5 and FIG. In addition, it is possible to know the image deterioration factor due to the interference convoluted by the image input device and its degree (S47). Based on this result, the image calibration procedure at the time of image input corresponding to the target image output apparatus is changed (S48).

  In the above-described process S41, the series of processes S42 to S48 after the calibration image 5a is read and input by the image input apparatus has been described. For example, the image input apparatus reads the entire print medium 5 and performs calibration. By specifying the output position of the image 5a, the normal image 5b output on the same print medium 5 can be simultaneously read and separated.

  The read and input normal image 5b is subjected to image degradation correction by the image calibration procedure changed in the process S48 by the processes S51 and S52 similar to the processes S32 and S33 shown in FIG. (S51) The corrected image data is stored in a recording medium such as a hard disk or a CD-ROM and is reused (S52).

  As described above, in the fourth embodiment, the calibration image 5a and the normal image 5b actually used are output on the same print medium 5, so that the calibration can be changed every time the input is performed. Thus, the image calibration procedure can be changed based on the image quality deterioration factor at the time of image input / output, regardless of the image output device and the output image.

  An image quality degradation determination method and an image calibration method according to the present invention use an image having a spatial frequency characteristic in which a power spectrum obtained by generalizing a natural image is approximately proportional to the reciprocal of the spatial frequency, and data of the spatial frequency characteristic. It is possible to determine the factor and its degree, and to use this determination result to improve the image quality in the image input device, to determine the image quality deterioration of the image input / output device, and to be useful for calibration.

A diagram in which the power of each point in the spatial frequency is made one-dimensional in the power spectrum of the natural image Diagram showing an example of calibration image A one-dimensional diagram of the power spectrum of the calibration image A one-dimensional diagram of the interference power spectrum in which noise (level 1 to 8) is convoluted in the calibration image A one-dimensional diagram of the interference power spectrum in which blur (level 1, 4, 8) is convoluted in the calibration image The flowchart which shows the process of the image quality degradation determination method in Embodiment 1 of this invention. The flowchart which shows the process which changes an image calibration procedure based on the result of the image quality degradation determination method in Embodiment 2 of this invention. Flowchart showing the normal image input procedure in the image input device after changing the image calibration procedure A diagram in which the power spectrum of interference convoluted by halftone processing (modulation) at the time of outputting a calibration image is made one-dimensional. The flowchart which shows the process which changes the image calibration procedure of the image input apparatus corresponding to the arbitrary image output apparatuses in Embodiment 3 of this invention. Flowchart showing a normal image input procedure corresponding to a specific image output device The figure which shows the image output on the same printing medium output from the image output device in Embodiment 4 of this invention, and input by the image input device. The flowchart which shows the process which changes the image calibration procedure of the image input apparatus corresponding to the arbitrary image output apparatuses in this Embodiment 4.

Explanation of symbols

5 Print medium 5a Calibration image 5b Normal image

Claims (3)

An image calibration method for an image input device that reads an image output from an image output device,
A calibration image having a spatial frequency characteristic whose power spectrum is approximately proportional to the reciprocal of the spatial frequency together with the main image is output on the same output medium by the image output device, and the output image is input from the image input device. Then, a one-dimensional power spectrum is obtained by comparing the input spatial frequency characteristic obtained from the input value of the calibration image and the spatial frequency characteristic of the calibration image, and the image is obtained based on the one-dimensional power spectrum. An image calibration method for an image input apparatus, wherein an image quality degradation factor and a degree of image quality degradation of the input apparatus are determined, and an image calibration procedure at the time of inputting the main image is changed based on the determination result.   The comparison between the input spatial frequency characteristic obtained from the input value of the image and the spatial frequency characteristic of the image is performed by the difference between the input spatial frequency characteristic and the spatial frequency characteristic of the image in Fourier space by Fourier transform processing. The image calibration method for an image input apparatus according to claim 1, wherein a dimensionalized power spectrum is acquired.   The determination of the image quality deterioration factor and the degree of image quality deterioration based on the one-dimensional power spectrum is performed by comparison with a known interference power spectrum obtained in advance by a deconvolution operation. 3. An image calibration method for an image input apparatus according to 2.