JP2006279162A - Method for inputting image and device for inputting image using the same - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device for inputting an image capable of improving an image quality by suppressing the increase of the ratio of noises even in the image at a low brightness level while preventing the generation of a tone jump, a pseudo-contour or the like. <P>SOLUTION: The problem is solved by constituting the noise ratios contained in each pixel of an image data so that dependencies on the brightness levels of pixels are reduced in an image input method for optically reading the image and converting the image into the image data. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image input method for photoelectrically converting a two-dimensional image into discrete image data, and an image input apparatus using the image input method. Specifically, the present invention is applied to a digital camera, an image scanner, and the like. The present invention relates to an image input method and an image input apparatus using the same.

Japanese Patent Application Laid-Open No. 2004-159087

  Conventionally, in an image input device such as a digital camera, an image is formed on a photoelectric conversion element such as a two-dimensional CCD (Charge Coupled Device) so as to obtain a discrete image signal corresponding to the intensity of light. It is configured. The image scanner is configured to obtain two-dimensional image data by moving the one-dimensional CCD along the sub-scanning direction of the image.

  In these image input devices such as a digital camera and an image scanner, when there is a so-called “dark noise” that occurs when a dark low-brightness image is converted into an image signal and input, the image quality deteriorates. . This dark noise is caused by thermal noise in a one-dimensional or two-dimensional CCD or electrical noise in a signal processing circuit, and particularly in the case of an image lacking absolute light quantity, such as an image of a night scene. As the shutter speed becomes slower, the amount of noise accumulated in the CCD increases, so that it becomes noticeable.

  In addition, the noise amount of the image input by the photoelectric conversion element such as the one-dimensional or two-dimensional CCD is almost constant without depending on the luminance level of the pixel, but is relatively included in the pixel at the low luminance level. As a result, the ratio of noise increases and the image quality deteriorates in a subjective impression on the viewer.

  Therefore, as a technique that can solve the problem of such image quality degradation, for example, as disclosed in Japanese Patent Application Laid-Open No. 2004-159087, noise is reduced by changing the size of the smoothing filter in accordance with the luminance level. There has already been proposed a technique configured to reduce this.

  A scanner system according to Japanese Patent Application Laid-Open No. 2004-159087 is a scanner system that captures a developed film with an imaging device and captures a video signal relating to an image on the film, and includes a predetermined size region from the video signal. Brightness detection means for detecting brightness information, film noise estimation means for estimating the granularity noise amount of the film for each area based on the brightness information detected by the brightness detection means, and estimation by the film noise estimation means Noise reduction means for reducing the noise of the video signal based on the granularity noise amount.

  However, the conventional technique has the following problems. That is, the scanner system according to Japanese Patent Application Laid-Open No. 2004-159087 includes a luminance detection unit that detects luminance information for each region of a predetermined size from a video signal, and the region based on the luminance information detected by the luminance detection unit. Film noise estimating means for estimating the granularity noise amount of the film every time, and noise reducing means for reducing the noise of the video signal based on the granularity noise amount estimated by the film noise estimating means. The noise reducing means is configured to reduce noise by changing the size of the smoothing filter based on the granularity noise amount.

  Therefore, in the case of the scanner system according to Japanese Patent Application Laid-Open No. 2004-159087, in the case of discrete image data, the size of the smoothing filter becomes discrete, and discontinuity occurs between the smoothing filters having different sizes. Has a problem.

  More specifically, in the case of the scanner system according to Japanese Patent Application Laid-Open No. 2004-159087, for example, the size of the smoothing filter is set to 3 × 3, 5 × 5, according to the average luminance around the pixel including the target pixel. If it is changed to 7 × 7 etc., discontinuity will occur in the vicinity where the smoothing filter changes, and in particular, in the case of an image whose gradation is continuously changed, such as a gradation image, the tone There is a possibility that jumps, pseudo contours, and the like may occur, which has a serious problem of reducing the image quality.

  Accordingly, the present invention has been made to solve the above-described problems of the prior art, and an object of the present invention is to suppress an increase in the ratio of noise even in an image having a low luminance level. It is possible to provide an image input device that can improve image quality by preventing occurrence of tone jumps, pseudo contours, and the like.

  The present invention is an image in which the noise rate included in each pixel constituting an image is made substantially constant by employing means for superimposing filters having different shapes in accordance with the average luminance level of surrounding pixels including the target pixel. An object of the present invention is to provide an input method and an image input apparatus using the same.

  In the present invention, when the signal value of the image data is S and the standard deviation of the signal value S is σs, the amount of noise included in the image data is defined by σs / S, and all of the low luminance to high luminance regions are defined. By reducing the amount of noise so that the amount of noise is constant at the signal value S, it is possible to realize an image quality with less noise.

  The noise reduction is realized, for example, by superimposing Gaussian filters having different weights according to the signal value of the image data. At this time, the coefficient a representing the spread of the Gaussian filter in the following equation 1 is set to be inversely proportional to the signal value S of the image data, so that the shape of the Gaussian filter is continuously smoothed according to the signal value S. Can be changed. In addition, since the wide-shaped filter is superimposed in the low-luminance portion, noise is further reduced, and in the high-luminance portion, the filter width is narrowed to be close to the original image.

  In general, when a wide filter is superimposed on the image data, there is a concern that the resolution deteriorates and the image quality deteriorates. The sensitivity is lower than that of the high-luminance frequency characteristic, and deterioration in resolution is hardly perceived.

  As described above, by applying the present invention, it is possible to realize a constant amount of noise in all signal values S from a low luminance to a high luminance region, noise can be reduced, and resolution is improved. It is possible to obtain a high-quality image without feeling any deterioration.

  Here, r is a distance from the filter center, S is an image signal value, and A is a constant.

  Specifically, the invention described in claim 1 is an image input method for optically reading an image and converting it into image data. The noise rate included in each pixel of the image data depends on the luminance level of the pixel. The image input method is characterized in that correction is performed so as to reduce the degree.

  According to a second aspect of the present invention, in an image input device that optically reads an image and converts it into image data, the noise rate included in each pixel of the image data is reduced in dependence on the luminance level of the pixel. An image input apparatus comprising an image correcting unit that corrects the image as described above.

  Furthermore, the invention described in claim 3 is characterized in that the image correction means performs correction by superimposing a filter whose shape is continuously changed according to an average luminance level of a pixel including a target pixel. The image input device according to claim 2.

  The invention described in claim 4 is characterized in that the image correcting means determines the shape of the filter in consideration of luminance values of peripheral pixels including the target pixel. An image input device. Here, considering the luminance value of the peripheral pixel including the target pixel means, for example, the case where the shape of the filter is determined using the luminance value of the peripheral pixel including the target pixel.

  Furthermore, the invention described in claim 5 is an image input device that optically reads an image and converts it into image data. Luminance level calculation means for obtaining an average luminance level of pixels including a target pixel of the image data; According to the average brightness level calculated by the brightness level calculation means, the calculation means for calculating the shape of the smoothing filter, and the smoothing filter calculated by the calculation means is superimposed to perform the smoothing process. This is an image input device.

  According to the present invention, it is possible to suppress an increase in the ratio of noise even in an image having a low luminance level, and to improve the image quality by preventing the occurrence of a tone jump or a pseudo contour. An image input device can be provided.

  Embodiments of the present invention will be described below with reference to the drawings.

First Embodiment FIG. 1 is a block diagram showing an image input apparatus to which an image input method according to a first embodiment of the present invention is applied.

  In this image input apparatus 1, as shown in FIG. 1, an image of a document in which a color or monochrome image is recorded is optically read by a scanner that scans a photoelectric conversion element such as a one-dimensional CCD (not shown) along the sub-scanning direction. Are read and converted into two-dimensional discrete image data, and the discrete image data 101 is stored in the image data storage unit 102. In the image input device 1, an image of a landscape or a person image is optically read by a photoelectric conversion element such as a two-dimensional CCD (not shown) using a digital camera or the like, and converted into discrete image data. Of course, the discrete image data may be stored in the image data storage unit 102.

  The uncorrected image data 101 stored in the image data storage unit 102 is sent to the average luminance level calculation unit 104 of the filter shape determination unit 103, and the target pixel of the image data 101 is processed by the average luminance level calculation unit 104. An average luminance level of a plurality of surrounding pixels including 105 is calculated.

  In this embodiment, as shown in FIG. 2, the average luminance level is averaged in the image of the region of 3 × 3 so that the averaged value becomes the average luminance value of the central pixel. It is configured. The shape determining unit 106 of the filter shape determining unit 103 is configured to determine the shape of the filter that smoothes the image data 101 using the average luminance value obtained by the average luminance level calculating unit 104 as described above. Yes.

  As the filter whose shape is determined by the filter shape determination unit 106, for example, a Gaussian filter represented by the following equation 1 is used. The shape parameter a that determines the shape of the Gaussian filter is determined as follows. First, a gray scale image is read and input by a scanner, and the relationship between the luminance value of the image and the amount of noise is obtained.

  Specifically, from the measured value of the gray scale image, the standard deviation of the luminance value at the average luminance value S1 is σ1, the standard deviation of the luminance value at the average luminance value S2 is σ2, and each filter shape When the determination coefficients are a1 and a2, the following expression (2) is satisfied so that the amount of noise is constant.

  At this time, in the present embodiment, the filter shape factor a when the gray scale density is high brightness (low density) of 0.08 is experimentally obtained, and a value of a = 0.1 is obtained as the filter shape coefficient a. Is adopted. The filter shape factor a is defined by the relationship a = A / S, and FIG. 3 shows the relationship between the output signal value S of the scanner and the filter shape determination coefficient a.

  FIG. 4 shows the filter shape when the filter shape determination coefficient a is 0.5, 1.0, 1.5, and 2.0. As is apparent from FIG. 4, when the filter shape determination coefficient a is as small as 0.5, the Gaussian filter has a sharp shape, whereas the filter shape determination coefficient a is as large as 2.0. As a result, the shape of the Gaussian filter is broad and wide.

  Coefficients are determined by using the filter thus obtained as a 9 × 9 digital filter. A known method can be used for determining the coefficient of the digital filter.

  FIG. 5 specifically shows the coefficients when the coefficients are determined as a 9 × 9 digital filter at a = 1.0 and a = 2.0.

  Further, as shown in FIG. 1, the filter superimposing unit 107 superimposes the digital filter created using the shape determination coefficient a obtained as described above on the image data 101, and the converted image is corrected. The image data 108 is stored in the image data storage unit 102.

  FIG. 6 shows the result of calculating the noise amount by superimposing a filter on the image actually read from the gray scale chart in order to confirm the effect of the present embodiment.

  As is apparent from FIG. 6, the amount of noise is substantially constant from a high luminance region where the output signal value is high to a low luminance region where the output signal value is low. It can be seen that an increase in the ratio can be suppressed.

  In the above embodiment, the shape of the filter to be used is continuously changed according to the luminance level of the image. Etc. can be prevented, and image quality can be improved.

Second Embodiment FIG. 7 shows a second embodiment of the present invention. The same reference numerals are given to the same parts as those in the first embodiment, and this second embodiment will be described. Then, the image correction means is configured to determine the shape of the filter in consideration of the differential value between the luminance value of the target pixel and the surrounding pixels.

  That is, in the second embodiment, as shown in FIG. 7, the filter shape is determined in consideration of the blur at the edge portion by using the Gaussian filter. For the calculation of the differential value in the differential value calculation unit 110, a 3 × 3 Sobel filter is used. As the filter shape determination coefficient a ′ taking into account the differential value, the following equation obtained experimentally is used.

a ′ = a−0.248 | Sob |
Here, a ′ is a shape determination coefficient after correction, a is a shape determination coefficient before correction, and Sob is a differential value obtained by superimposing a Sobel filter.

  As described above, in the second embodiment, when the differential value is large using the differential value of the image data 101, that is, in an image with a clear edge portion, the predetermined value 0 than the original shape determination coefficient a. By using a shape determination coefficient a ′ that is smaller by .248 | Sob |, blurring at the edge portion is suppressed.

  Therefore, in the second embodiment, the amount of noise can be made substantially constant from a high luminance region with a high output signal value to a low luminance region with a low output signal value. It is possible to suppress an increase in the noise ratio, and it is possible to suppress blurring at the edge portion due to the use of a Gaussian filter.

  Other configurations and operations are the same as those of the first embodiment, and thus description thereof is omitted.

Experimental Example A subjective evaluation experiment was performed to confirm the effect of the present invention. For the evaluation, night views and female portrait images taken with a digital camera were used. For each image, there is no correction, the correction according to the present invention is performed, and the three types of patterns when the correction is performed with the coefficient considering the blur at the edge portion according to the present invention are good, ordinary, and bad. Three levels of scores were assigned to each image, and the average value of the evaluation values of the subjects was used as the score of the image. There were 12 subjects. The results are shown in FIG.

  As is apparent from FIG. 8, the subjective evaluation value is improved as a result of the processing according to the present invention for both the night view and the human image, and the effect of the present invention has been confirmed.

  According to the present invention, in an image input device that optically reads an image and converts it into discrete image data, the noise rate included in each pixel of the discrete image data is constant without depending on the luminance level of the pixel. By including the image correcting means for correcting the dark noise, the dark part noise was successfully reduced. In the present invention, by changing the shape of the Gaussian filter depending on the local average luminance value of the image, it is possible to obtain an image free from noise from the highlight to the shadow portion. Further, by using the Gaussian filter, the shape of the filter is determined in consideration of the influence of the blur at the edge portion, so that an image with the minimum influence of the blur is realized. The result of the subjective evaluation experiment also confirmed the effect of the processing according to the present invention.

FIG. 1 is a block diagram showing an image input apparatus according to Embodiment 1 of the present invention. FIG. 2 is an explanatory diagram showing how to obtain the luminance value of the target pixel. FIG. 3 is a graph showing the relationship between the output signal value and the filter shape determination coefficient. FIG. 4 is a graph showing different shapes of the filter. FIG. 5 is a chart showing filter coefficients. FIG. 6 is a graph showing the relationship between the output signal value and the amount of noise. Between filter and filter shape determination coefficient FIG. 7 is a block diagram showing an image input apparatus according to Embodiment 2 of the present invention. FIG. 8 is a graph showing a subjective evaluation result for confirming the effect of the present invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1: Image input device 101: Image data 102: Image data storage part 103: Filter shape determination part 104: Average brightness level calculation part 106: Shape determination part

Claims (5)

In an image input method for optically reading an image and converting it into image data,
An image input method, wherein a noise rate included in each pixel of the image data is corrected so as to reduce dependency on a luminance level of the pixel. In an image input device that optically reads an image and converts it into image data,
An image input apparatus comprising: an image correction unit that corrects a noise rate included in each pixel of the image data so as to reduce dependence on a luminance level of the pixel. The image input device according to claim 2, wherein the image correction unit performs correction by superimposing a filter whose shape is continuously changed according to an average luminance level of a pixel including a target pixel. The image input device according to claim 2, wherein the image correction unit determines the shape of the filter in consideration of luminance values of peripheral pixels including the target pixel. In an image input device that optically reads an image and converts it into image data,
A luminance level calculating means for obtaining an average luminance level of pixels including the target pixel of the image data;
A computing means for computing the shape of the smoothing filter in accordance with the average brightness level calculated by the brightness level calculating means;
An image input apparatus for performing a smoothing process by superimposing a smoothing filter calculated by the calculation means.

Images