JP2008283724A - Method, apparatus and program for image processing - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To quantify the sense of contrast perceived by a viewer of an image, and to carry out adequate image processing on image data based on the sense of contrast. <P>SOLUTION: Contrast-sense quantification means 1 generates unsharp image data of image data S0 and then generates a histogram of the unsharp image data. Since the histogram of the image data S0 includes lightness information of details of the image, a distribution width thereof does not represent the contrast perceived by the viewer of the image represented by the image data S0 as a whole. However, since the histogram of the unsharp image data excludes information of the details in the image, a distribution width of the unsharp image data represents the contrast of the overall image. The distribution width of the histogram of the unsharp image data is then found as the sense of contrast C0 and inputted to processing means 2. In the processing means 2, tone conversion processing is carried out on the image data S0 by changing a tone conversion LUT based on the sense of contrast C0, and processed image data S1 are obtained. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention provides an image processing method and apparatus for quantifying a sense of contrast actually received from an image by a person who has viewed the image, and further performing image processing on image data based on the quantified contrast feeling. The present invention relates to a program for causing a computer to execute a processing method.

  Digital image data acquired by a digital camera or digital image data obtained by reading an image recorded on a film is reproduced as a hard copy such as a print or as a soft copy on a display. As described above, when reproducing digital image data, various image processing such as gradation processing and frequency processing are performed on the image in order to have the same high-quality image as a photograph printed from a negative film. Applies to data.

  For example, a histogram of image data is created, the contrast of the image represented by the image data is obtained from the distribution width of the histogram, and a gradation curve for converting the gradation of the image data is corrected based on this contrast. Accordingly, various image processing methods have been proposed for converting image data so that gradation is not crushed and noise is not noticeable (see, for example, Patent Document 1). The contrast here refers to the ratio of the dark part to the bright part in the image. Therefore, the contrast of the image can be determined from the distribution width of the histogram, such that a histogram having a broad distribution has a contrast and a histogram having a narrow histogram has no contrast. For example, an image obtained by shooting under clear sky is a histogram with a wide distribution range that reflects the light and darkness from the sun to the shade, and an image obtained by shooting under cloudy weather is sunny and shaded. The histogram is distributed with a narrow width that is difficult to distinguish.

In addition, an image processing method that expresses human senses such as sharpness and graininess when observing an image as a numerical value and changes the content of the image processing based on the numerical value to obtain a preferable image for human beings. Has also been proposed (see Patent Document 2).
JP-A-6-253176 JP-A-7-193766

  As described above, the histogram created from the image data has a complicated shape because it includes various information in all subjects and image details included in the image. Since the contrast information perceived when actually observing the image is buried, the sense of contrast perceived by humans is not necessarily reflected. For example, in the case of an image with a human face as the subject, the person who observes this image perceives the contrast only in the face portion, and for subjects other than the face included in the image, the contrast received from the image Are not perceived. However, since the histogram created from the image data includes information about subjects other than the face, this histogram does not reflect the contrast perceived by the person observing the image. Therefore, if image processing is performed on image data based on such a histogram, it is not always possible to obtain a processed image desired by a person who observes the image.

  Also, the contrast perceived by humans varies depending on the contrast between the vivid colors and the non-vivid colors included in the image. For example, an image containing a lot of bright colors is perceived as having a sense of contrast, whereas an image containing a lot of non-brilliant colors is perceived as having no sense of contrast. As described above, since the contrast differs depending on the color included in the image, it is necessary to perform image processing in consideration of the color of the image.

  The present invention has been made in view of the above circumstances, and by actually observing an image, a sense of contrast perceived by a human is quantified as a sense of contrast, and an image is appropriately displayed for the image based on the sense of contrast. An object of the present invention is to provide an image processing method and apparatus capable of performing processing, and a program for causing a computer to execute image processing.

  Another object of the present invention is to provide an image processing method and apparatus capable of appropriately performing image processing on an image using color information of the image, and a program for causing a computer to execute image processing. To do.

  When a human observes an image and determines the contrast of the image, not only the contrast of all the subjects included in the image but also the overall contrast of the entire image, the bright and dark areas in the image The contrast of the image is determined based on information that is not reflected in the histogram, such as the distribution of the image, and the distribution of brightness of only the subject of interest. The present invention has been made paying attention to this point.

That is, in the first image processing method according to the present invention, the conversion means calculates color data representing the color information of the image represented by the image data from the image data,
A blur image creating means creates blur image data of the color data,
A histogram creating means creates a histogram representing a two-dimensional frequency distribution of the blurred image data;
The contrast feeling calculating means calculates the distribution area of the histogram as a contrast feeling,
The processing means performs image processing for changing the luminance information of the image on the image data based on the sense of contrast.

A first image processing apparatus according to the present invention comprises: conversion means for calculating color data representing color information of an image represented by the image data from the image data;
A blurred image creating means for creating blurred image data of the color data;
A histogram creating means for creating a histogram representing a two-dimensional frequency distribution of the blurred image data;
Contrast feeling calculating means for calculating the distribution area of the histogram as a contrast feeling;
And a processing means for performing image processing for changing the luminance information of the image on the image data based on the sense of contrast.

  The second image processing method according to the present invention is characterized in that the contrast feeling of an image represented by the image data is quantified based on the image data.

  Here, "contrast feeling" is not directly reflected in the histogram of the image itself, such as the overall brightness difference of the entire image, the distribution of bright and dark parts in the image, the distribution of light and darkness in the subject of interest, This is a general subjective sensation regarding contrast actually received from an image by a person who has observed the image. Specifically, a histogram of the blurred image data of the image data, a bright portion of the blurred image represented by the blurred image data, and / or Alternatively, it can be quantified based on the position information of the dark part, the histogram obtained from the multi-resolution image data for each frequency band obtained by converting the image data to multi-resolution.

  In order to create a histogram from the blurred image data, the histogram may be created from the blurred image data itself. For example, when the image data has 8-bit (256) information, the blurred image data is converted into, for example, 32 values. , 16-valued, 8-valued, etc., and a histogram may be created from the blurred image data that has been 32-valued.

  Further, luminance data and color data representing luminance information and color information of the image are obtained from the image data, and luminance blur image data and / or color blur image data which are blur image data of the luminance data and / or color data are created, A brightness histogram and / or a color histogram may be created from the brightness blur image data and / or the color blur image data, and the sense of contrast may be quantified based on the brightness histogram and / or the color histogram. “Color information” refers to information representing the vividness of colors included in an image.

  When color blur image data is created, a color histogram representing a two-dimensional frequency distribution of the color blur image data may be created.

  Furthermore, as the “position information of the bright part and / or dark part”, for example, the standard deviation of the distance from the center of the image to the bright part and / or dark part can be used.

  Furthermore, “based on a histogram or the like obtained from multi-resolution image data” specifically means that when multi-resolution image data of high frequency band, medium frequency band and low frequency band is created from image data, Roughness distribution of the image obtained from the low frequency band image represented by the low frequency band resolution image data, medium frequency band image or high frequency band image represented by the resolution image data of the medium frequency band or high frequency band This is based on the histogram and the like.

  Further, luminance data and color data representing the luminance information and color information of the image are obtained from the image data, the luminance data and / or the color data are converted into multi-resolution, and the luminance multi-resolution image data for each of a plurality of frequency bands and / or Color multi-resolution image data is obtained, a luminance histogram and / or a color histogram, which is a histogram of each luminance multi-resolution image data and / or each color multi-resolution image data, is created, and a contrast feeling is based on each luminance histogram and / or each color histogram. May be quantified.

  In the second image processing method according to the present invention, it is preferable to perform image processing on the image data based on the sense of contrast.

  In this case, the image processing is preferably at least one of gradation changing processing, frequency enhancement processing, AE processing, and saturation changing processing.

  A third image processing method according to the present invention is characterized in that image processing for changing luminance information of the image is performed on the image data based on color information of the image represented by the image data from the image data. To do.

In the third image processing method according to the present invention, color data representing color information of the image is obtained from the image data,
Create blurred image data of the color data,
Create a histogram of the blurred image data,
The image processing is preferably performed on the image data based on the histogram.

  In this case, it is preferable to create a histogram representing a two-dimensional frequency distribution of the blurred image data.

In the third image processing method according to the present invention, color data representing color information of the image is obtained from the image data,
The color data is converted into multi-resolution to obtain multi-resolution image data for each of a plurality of frequency bands,
Create a histogram of the multi-resolution data of the lowest frequency band among the multi-resolution image data,
The image processing is preferably performed on the image data based on the histogram.

  The second image processing apparatus according to the present invention is characterized by comprising contrast quantification means for quantifying the contrast feeling of the image represented by the image data based on the image data.

In the second image processing apparatus according to the present invention, the contrast quantification unit includes a blur image data creation unit that creates blur image data of the image data,
A histogram creating means for creating a histogram of the blurred image data;
It is preferable to include a quantification unit that quantifies the contrast feeling based on the histogram.

Further, in the second image processing apparatus according to the present invention, the contrast quantification means includes conversion means for obtaining luminance data and color data representing luminance information and color information of the image from the image data,
Blur image data creation means for creating brightness blur image data and / or color blur image data which is the blur data of the brightness data and / or the color data;
A histogram creation means for creating a brightness histogram and / or a color histogram which is a histogram of the brightness blur image data and / or the color blur image data;
It is preferable to include a quantification unit that quantifies the contrast feeling based on the luminance histogram and / or the color histogram.

  In this case, when the blur image data creation unit creates the color blur image data, the histogram creation unit is preferably a unit that creates a color histogram representing a two-dimensional frequency distribution of the color blur image data. .

Further, the contrast sensation quantification means includes a blur image data creation means for creating blur image data of the image data,
It is preferable that the apparatus further includes a quantification unit that quantifies the contrast feeling based on position information of a bright part and / or a dark part in a blurred image represented by the blurred image data.

Further, the contrast sensation quantifying means converts the image data into multi-resolution to obtain multi-resolution image data for each of a plurality of frequency bands; and
A histogram creating means for creating a histogram of each multi-resolution image data;
It is preferable to include a quantification unit that quantifies the contrast feeling based on the histograms.

Furthermore, the contrast quantification means includes conversion means for obtaining luminance data and color data representing luminance information and color information of the image from the image data,
Multi-resolution conversion means for converting the luminance data and / or the color data into multi-resolution to obtain luminance multi-resolution image data and / or color multi-resolution image data for each of a plurality of frequency bands;
Histogram creation means for creating a luminance histogram and / or a color histogram that is a histogram of each luminance multi-resolution image data and / or each color multi-resolution image data;
It is preferable to include a quantification unit that quantifies the contrast feeling based on each luminance histogram and / or each color histogram.

  The second image processing apparatus according to the present invention preferably further includes processing means for performing image processing on the image data based on the contrast feeling.

  In this case, the processing means is preferably means for performing at least one of gradation changing processing, frequency enhancement processing, AE processing, and saturation changing processing as the image processing.

  According to a third image processing apparatus of the present invention, the image data is subjected to image processing for changing the luminance information of the image based on the color information of the image represented by the image data. To do.

In the third image processing apparatus according to the present invention, conversion means for obtaining color data representing color information of the image from the image data;
A blurred image data creating means for creating blurred image data of the color data;
A histogram creating means for creating a histogram of the blurred image data;
Preferably, the image processing apparatus includes processing means for performing the image processing on the image data based on the histogram.

  In this case, the histogram creation means is preferably means for creating a histogram representing a two-dimensional frequency distribution of the blurred image data.

In the third image processing apparatus according to the present invention, conversion means for obtaining color data representing color information of the image from the image data;
Multi-resolution conversion means for converting the color data into multi-resolution to obtain multi-resolution image data for each of a plurality of frequency bands;
Histogram creating means for creating a histogram of the multi-resolution image data in the lowest frequency band among the multi-resolution image data,
Preferably, the image processing apparatus includes processing means for performing the image processing on the image data based on the histogram.

  The first to third image processing methods according to the present invention may be provided as a program for causing a computer to execute.

  According to the present invention, since the contrast feeling of the image represented by the image data is quantified, the image is not a contrast including various information of the entire image like the contrast obtained from the histogram of the image itself. The subjective contrast received from an image when a person actually observes the image, such as the overall brightness difference of the image, the distribution of light and dark areas in the image, the distribution of light and darkness in the subject of interest, and the color information contained in the image Sense can be quantified.

  In addition, since the image represented by the image data includes a great deal of information as well as information perceived by the person observing the image, the histogram created from the image data includes a large image. The light and dark information is buried. On the other hand, by creating blurred image data from the image data, the image represented by the blurred image data does not include detailed pixel value changes of the subject unlike the original image data. It clearly represents the global contrast of the entire image that is actually perceived when observed. Therefore, based on the blurred image data of the image data, or the histogram of the luminance data and / or color data obtained from the image data, the sense of contrast received from the image when actually observing the image is satisfactorily quantified. be able to.

  Further, since the position information of the bright part and / or the dark part in the blurred image represents the position of the bright subject and / or the dark subject in the image, the distribution state of the light and darkness in the image is based on this position information. Can be obtained as a sense of contrast.

  Furthermore, when the image data is converted into multi-resolution image data and a histogram of each resolution image data is created, the histogram of the resolution image data in the low frequency band shows the light and dark distribution of the entire image, similar to the histogram of the blurred image data, The histogram of the resolution image data in the middle and high frequency band represents the amplitude of the frequency component corresponding to the frequency band. For example, the darkness of the face due to the nose and eye depression, the shadow due to the building and subject, etc., due to the medium frequency component, the tree branches and the fineness of the flowers, the pattern of the person's clothes, the texture, and the boundaries (edges) between the objects are high. It is composed of frequency components. For this reason, the distribution width of the histogram of the resolution image data in the middle and high frequency band becomes larger as the image having local contrast in these images. Therefore, the overall contrast feeling of the image data can be quantified by the histogram of the resolution image data in the low frequency band, and the local contrast feeling in the image can be quantified based on the histogram of the resolution image data in the middle and high frequency band. The image can be quantified, whereby not only the overall light and dark distribution of the image but also the local light and dark distribution can be obtained as a contrast feeling.

  On the other hand, when luminance data and color data are obtained from image data, and the luminance data and / or color data are converted into multi-resolution image data to create a luminance histogram and / or color histogram that is a histogram of each resolution image data, The luminance histogram in the frequency band represents the light / dark distribution of the entire image, and the luminance histogram in the middle / high frequency band represents the amplitude of the frequency component corresponding to the frequency band. Therefore, the overall contrast feeling of the image data can be quantified based on the luminance histogram in the low frequency band, and the local contrast feeling in the image can be quantified based on the luminance histogram in the middle and high frequency band. Can do.

  The color histogram in the low frequency band represents the saturation distribution of the entire image, and the color histogram in the middle and high frequency band represents the saturation distribution according to the frequency band. Therefore, it is possible to quantify the overall contrast of the image data based on the color of the image based on the color histogram in the low frequency band, and further, local contrast in the image based on the color histogram in the middle and high frequency band The feeling can be quantified.

  Furthermore, by performing predetermined image processing on the image data based on the obtained sense of contrast, processed image data reflecting the sense of contrast perceived by the person observing the image can be obtained.

  In addition, by performing image processing that changes luminance information on the image data based on the color information of the image represented by the image data from the image data, a person who observes the image perceives the contrast feeling perceived from the color of the image. Processed image data reflecting the above can be obtained.

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

  FIG. 1 is a schematic block diagram showing the configuration of an image processing apparatus according to an embodiment of the present invention. As shown in FIG. 1, the image processing apparatus according to the present embodiment is quantified by a contrast sensation quantifying unit 1 that quantifies a contrast sensation C0 in an image represented by image data S0 and a contrast sensation quantifying unit 1. Based on the contrast C0, the processing means 2 that performs image processing on the image data S0 to obtain the processed image data S1, and the output of the printer, CRT monitor, etc. that outputs the processed image data S1 as a visible image Means 3 are provided.

  FIG. 2 is a schematic block diagram showing a specific configuration of the contrast quantification means 1. The contrast quantification means 1 shown in FIG. 2 will be described as the first embodiment. As shown in FIG. 2, the contrast quantification unit 1 according to the first embodiment includes a blur image creation unit 11 that creates blur image data Sus of image data S0, and a histogram that creates a histogram Hus of blur image data Sus. A creation unit 12 and a quantification unit 13 for quantifying and obtaining the contrast feeling C0 of the image represented by the image data S0 based on the histogram Hus are provided.

  In the contrast sensation quantifying means 1 shown in FIG. 2, the contrast sensation C0 is quantified and determined as follows. First, the blurred image creation unit 11 creates blurred image data Sus of the image data S0. The generation of the blurred image data Sus is performed, for example, by performing a filtering process using a blur mask filter on the image data S0. An example of the image represented by the image data S0 and the blurred image represented by the blurred image data Sus is shown in FIG. In FIG. 3, the images represented by the image data S0 and the blurred image data Sus are given the corresponding symbols S0 and Sus, respectively. The blurred image data Sus is such that a frequency band of several percent remains with respect to the Nyquist frequency of the image data S0. Specifically, 0.5 to 3 cycles in the image represented by the image data S0. It represents a frequency component of about / cm.

  Next, the histogram creation means 12 creates a histogram Hus of the blurred image data Sus. FIG. 4 is a diagram showing the histogram Hus of the blurred image data Sus together with the histogram H0 of the image data S0. In FIG. 4, the pixel value is normalized to 0-100. As shown in FIG. 4, the histogram H0 has a complicated shape because it includes not only the light and dark distribution of the entire image but also various information on all subjects and image details included in the image. The contrast or contrast obtained from the histogram H0, that is, the contrast over a very wide range of 0-100. FIG. 5 is a diagram showing image data S0 ′ different from the image data S0 and histograms H0 ′ and Hus ′ of the blurred image data Sus ′ created therefrom. Comparing FIG. 4 and FIG. 5, the shape of the histogram is clearly different, but the contrast obtained from the histogram H0 ′ in FIG. 5 is in a wide range of 0-100, similar to the contrast obtained from the histogram H0 in FIG. Therefore, there is no difference in contrast between images.

  On the other hand, the distribution width of the histogram Hus is narrower than that of the histogram H0, which is 27-92. However, detailed information included in the image is removed, and this distribution width is a global light / dark difference of the entire image, that is, It represents the contrast. Further, the contrast obtained from the histogram H0 'shown in FIG. 5 is the same as that shown in FIG. 4, but when the histogram Hus' and the histogram Hus are compared, the distribution width of the histogram Hus' is 17-75. Since the distribution width and the distribution position are clearly different from each other, a difference in contrast appears depending on the image. Here, when a human observes an image, the contrast is determined by first observing the entire image, not a detailed portion in the image. Therefore, in the first embodiment, the quantification unit 13 obtains the distribution width of the histogram Hus of the blurred image data Sus, that is, the difference w between the maximum value Husmax and the minimum value Husmin of the histogram Hus, and this is calculated by the human. A contrast feeling C0 representing the contrast of the entire image perceived when observing.

  The processing means 2 performs image processing on the image data S0 based on the contrast feeling C0 quantified by the contrast feeling quantifying means 1. First, the contrast sense C0, that is, the difference w between the maximum value Husmax and the minimum value Husmin of the histogram Hus is compared with a preset threshold Th1 to obtain the contrast type of the image represented by the image data S0. Here, as shown in FIGS. 4 and 5, the threshold value Th <b> 1 is set to, for example, about 50 when the pixel values of the entire histogram are distributed in the range of 1-100, but is not limited thereto. When C0 ≧ Th1, the image represented by the image data S0 is determined as a high contrast image, and when C0 <Th1, it is determined as a low contrast image. In this case, two threshold values Th1 and Th2 are set such that 0 <Th2 <Th1 <100. When C0> Th1, a high contrast image is set, when Th2 ≦ C0 ≦ Th1, a standard image is set, and C0 < In the case of Th2, a discrimination such as a low contrast image may be performed. In this case, Th1 is set to about 80 and Th2 is set to about 40, but the present invention is not limited to this.

  When the contrast type of the image is obtained in this way, a gradation conversion LUT prepared in advance according to the contrast type is selected, and gradation conversion is performed on the image data S0 by the selected gradation conversion LUT. Processing is performed. FIG. 6 is a diagram showing the gradation conversion LUT. In the present embodiment, five gradation conversion LUTs LUT1 to LUT5 are prepared. When it is determined as a high-contrast image, it is determined as LUT5. When it is determined as a low-contrast image, it is determined as LUT1 and a standard image. In this case, image processing for obtaining processed image data S1 by converting the gradation of the image data S0 using the gradation conversion LUT of LUT3 is performed. Note that the gradation conversion LUT of LUT1 to LUT5 may be selected in accordance with the contrast C0, that is, the value of the difference w.

  The gradation curve is expressed by a function such as, for example, yout = a · yin + b (yout: output, yin: input), and a and b are expressed in accordance with the contrast type of the image or the sense of contrast C0 (value of difference w). The gradation curve may be set by changing parameters.

  Next, the operation of the first embodiment will be described.

  FIG. 7 is a flowchart showing the operation of the first embodiment. In the flowchart shown in FIG. 7, it is assumed that the contrast feeling C0 is compared with two threshold values Th1 and Th2. First, the blurred image creating unit 11 of the contrast quantification unit 1 creates the blurred image data Sus of the image data S0 (step S1), and the histogram creating unit 12 further creates the histogram Hus of the blurred image data Sus. (Step S2). The quantifying means 13 quantifies and obtains the contrast feeling C0 based on the histogram Hus (step S3). Then, the obtained contrast feeling C0 is input to the processing means 2, and it is first determined whether or not C0> Th1 (step S4). When step S4 is affirmed, the gradation conversion LUT5 is selected as a high-contrast image (step S5), and based on this, gradation conversion processing is performed on the image data S0 (step S6). Processed image data S1 is obtained. The obtained processed image data S1 is output as a visible image by the output means 3 (step S7).

  On the other hand, if step S4 is negative, it is determined whether C0 <Th2 is satisfied (step S8). If step S8 is positive, the gradation conversion LUT1 is selected as a low contrast image (step S9), and gradation conversion processing is performed on the image data S0 based on this (step S6). . Further, if step S8 is negative, the gradation conversion LUT3 is selected as a standard image satisfying Th2 ≦ C0 ≦ Th1 (step S10), and gradation conversion processing is performed on the image data S0 based on this. Is applied (step S6).

  Here, since the image represented by the image data S0 includes not only information perceived by the person observing the image but also a great deal of information, as shown in FIG. 4 and FIG. In the histogram H0 created from the data S0, the light and dark information of the entire image is buried. On the other hand, since the image represented by the blurred image data Sus of the image data S0 does not include a change in the detailed pixel value of the subject unlike the image data S0, it is actually perceived when a human observes the image. It represents the global contrast of the entire image. Therefore, based on the histogram Hus of the blurred image data Sus, the contrast feeling C0 received from the image when actually observing the image can be quantified satisfactorily, and the image data S0 is based on the contrast feeling C0. By applying gradation conversion processing to the image, processed image data S1 representing an image reflecting the contrast C0 perceived by a person observing the image can be obtained.

  In the first embodiment, the histogram H0 is created from the blurred image data Sus itself. For example, when the blurred image data Sus has an 8-bit (0-255) data value, The blurred image data Sus may be converted into 16 values, and a histogram may be created from the 16-valued blurred image data Sus. At this time, since the blurred image data Sus does not include detailed information in the image unlike the image data S0, even if the histogram Hus is created after 16-value conversion, the distribution width is as shown in FIG. Since it does not change so much as compared with the case where it is created from 8-bit blurred image data Sus, the obtained contrast feeling C0 is not so different before and after the 16-value conversion. Further, since the data amount of the pixel value is smaller when the 16-value is used, a histogram can be easily created. Therefore, processing can be performed at high speed by creating a histogram after converting the blurred image data Sus into 16 values. In this case, the data value of the blurred image data Sus is 16-valued. However, if the value is smaller than 8 bits, for example, an 8-value or 32-value data is created and a histogram is created. Also good.

  In the first embodiment, the difference w between the maximum value Husmax and the minimum value Husmin of the histogram Hus is set as the contrast feeling C0. For example, the 10% value of (Husmax−Husmin) from the maximum value Husmax of the histogram Hus. A value of a position where the 10% value of (Husmax−Husmin) is large may be obtained from the value of the small position and the minimum value Husmin of the histogram Hus, and the difference between the values at these positions may be obtained as the contrast C0.

  Further, in the first embodiment, the blurred image data Sus is created from the image data S0, and the contrast feeling C0 is quantified from the histogram Hus of the blurred image data Sus. However, the image data S0 is represented by the image data S0. It is also possible to convert the luminance information and color data representing the luminance information and color information of the image to be obtained, and to quantify the sense of contrast C0 from the histogram of the luminance data and / or color data. Hereinafter, this will be described as a second embodiment.

FIG. 9 is a schematic block diagram showing the configuration of the second embodiment of the contrast quantification means 1. As shown in FIG. 9, the contrast quantification means 1 according to the second embodiment receives image data S0 in addition to the blurred image creation means 11, the histogram creation means 12, and the quantification means 13 in the first embodiment. Conversion means 15 for converting into luminance data L ^* and color data C ^* is provided, and in the blurred image creation means 11, luminance blurred image data Lus or color blurred image data Cus, which is blurred image data of luminance data L ^* or color data C ^*. And the histogram creation means 12 creates a brightness histogram HLus that is a histogram of the brightness blur image data Lus or a color histogram HCus that is a histogram of the color blur image data Cus.

In the conversion means 15, the image data S0 is converted into luminance data L ^* and color data C ^* as follows. In the second embodiment, the image data S0 is ITU-R BT. It is assumed that RGB color data R0, G0, B0 conforming to 709 (REC. 709) is included. In the conversion means 15, the color data R0, G0, B0 constituting the image data S0 are converted into CIE1931 tristimulus values X, Y, Z based on the following formulas (1) to (3).

Pr = R0 / 255
Pg = G0 / 255 (1)
Pb = B0 / 255

R1 '= ((Pr + 0.099) /1.099) ^2.222
G1 '= ((Pg + 0.099) /1.099) ^2.222 (Pr, Pg, Pb ≧ 0.081) (2)
B1 '= ((Pb + 0.099) /1.099) ^2.222

R1 ′ = Pr / 4.5
G1 ′ = Pg / 4.5 (Pr, Pg, Pb <0.081) (2 ′)
B1 '= Pb / 4.5

X R0 '
Y = | A | ・ G0 ′ (3)
Z B0 '
Here, the matrix | A | is a matrix for converting the color data R0 ′, G0 ′, and B0 ′ into the tristimulus values X, Y, and Z. For example, the following values can be used.

0.4124 0.3576 0.1805
| A | = 0.2126 0.7152 0.0722 (4)
0.0193 0.1192 1.0571
Instead of the matrix | A |, the tristimulus values X, Y, and Z may be obtained by a lookup table.

Next, CIE 1976 L ^* , a ^* , and b ^* are obtained from the tristimulus values X, Y, and Z by the following formulas (5) to (7).

a ^* = 500 {f (X / Xn) -f (Y / Yn)} (5)
b ^* = 200 {f (Y / Yn) -f (Z / Zn)} (6)
L ^* = 116 (Y / Yn) ^1/3 -16 (when Y / Yn> 0.008856) (7)
L ^* = 903.25 (Y / Yn) (when Y / Yn ≦ 0.008856)
here,
When X / Xn, Y / Yn, Z / Zn> 0.008856 f (a / an) = (a / an) ^1/3 (a = X, Y, Z)
When X / Xn, Y / Yn, Z / Zn ≦ 0.008856 f (a / an) = 7.787 (a / an) +16/116
Xn, Yn, and Zn are tristimulus values for white and are tristimulus values corresponding to CIE-D65 (light source having a color temperature of 6500K). Further, the saturation C ^* is obtained by the following equation (8).

C ^* = (a ^{* 2} + b ^{* 2} ) ^1/2 (8)
Then, L ^* is output as luminance data, and C ^* is output as color data.

From the brightness data L ^* or the color data C ^* , the blur image creation means 11 performs the brightness blur image data Lus or the color blur image, which is the blur image data of the brightness data L ^* or the color data C ^* , as in the first embodiment. Data Cus is created. Note that the color data C ^* is less blurred than the luminance data L ^{* in} consideration of not only the global color change but also the influence of medium frequency components such as fine flowers. Also good. Specifically, it is preferable that the color blurred image data Cus represents a frequency component of about 0.5 to 10 cycles / mm in the image represented by the image data S0.

  In the histogram creating means 12, as in the first embodiment, the brightness histogram HLus is created from the brightness blurred image data Lus, or the color histogram HCus is created from the color blurred image data Cus.

  Then, the quantification means 13 obtains the distribution width of the luminance histogram HLus or the color histogram HCus, and outputs this as the contrast feeling C0.

  Based on the contrast feeling C0 thus obtained, the processing means 2 performs image processing on the image data S0. Specifically, as in the first embodiment, the type of contrast of the image represented by the image data S0 is determined based on the sense of contrast C0, and gradation conversion prepared in advance according to the determination result A LUT is selected, and gradation conversion processing is performed on the image data S0 by the selected gradation conversion LUT.

In the second embodiment, when the contrast feeling C0 is obtained from the color data C ^* and the contrast type obtained from the contrast feeling C0 is a low contrast image, the image represented by the image data S0. You may make it perform the saturation change process which raises the saturation of. Specifically, the saturation is improved by multiplying the saturation C ^* obtained by the above equation (8) by the enhancement coefficient αc. Note that the value of the enhancement coefficient αc is preferably about 1.2, but is not limited thereto. Further, the gradation conversion process and the saturation change process may be performed simultaneously.

Further, in the case of improving the saturation, not only the uniform enhancement coefficient αc is multiplied to the entire image represented by the image data S0, but the saturation is further improved in the low saturation portion in the image. The enhancement coefficient αc may be set as a function of saturation. The enhancement coefficient αc may be changed according to the hue angle H (= tan ⁻¹ (b ^* / a ^* )).

In the second embodiment, the conversion means 15 obtains the saturation C ^* by the above equation (8) and uses this as the color data, but a ^* and b obtained by the equations (5) and (7). ^{* May} be color data. In this case, the blurred image creation means 11 obtains a ^* and b ^* color blur image data aus and bus, and the histogram creation means 12 creates a two-dimensional histogram Hab from the color blur image data aus and bus. The FIG. 10 is a diagram showing an example of a two-dimensional histogram Hab. In FIG. 10, the distance from the origin represents the saturation, and the brighter the color, the more away from the origin. For this reason, if there are many vivid colors in the image represented by the image data S0, the distribution of the two-dimensional histogram Hab will spread. For example, in FIG. 10 (a) and FIG. 10 (b), the distribution of the two-dimensional histogram Hab is wider in FIG. 10 (b), so the two-dimensional histogram Hab shown in FIG. 10 (b) was obtained. As a result, the image contains more vivid colors.

  Therefore, the distribution area Ac of the two-dimensional histogram Hab can be obtained, and this distribution area Ac can be used as the contrast feeling C0. Then, based on the sense of contrast C0, the contrast type of the image represented by the image data S0 is determined in the same manner as in the first embodiment, and gradation processing, saturation enhancement processing, etc. are performed on the image data S0. Image processing can be performed.

In the second embodiment, the number of pixels P where the color data C ^* is equal to or greater than a predetermined threshold is counted, and the ratio R = P / of the number of pixels and the total number of pixels Pall of the image represented by the image data S0. Pall may be obtained and this may be obtained as the contrast feeling C0. Specifically, in the aus-bus plane shown in FIG. 10, a circular area having a predetermined radius (having a value corresponding to a predetermined threshold) around the origin is set, and the number of pixels not included in the circular area Is the number of pixels P, a ratio R of the number of pixels P to the total number of pixels Pall is obtained, and this is used as a contrast feeling C0.

  In the first embodiment, the histogram Hus of the blurred image data Sus of the image data S0 is created, and the contrast feeling C0 is quantified based on the histogram Hus. You may obtain | require and quantify the distribution state of a dark part as the contrast feeling C0. Hereinafter, this will be described as a third embodiment. FIG. 11 is a schematic block diagram showing the configuration of the third embodiment of the contrast quantification means 1. As shown in FIG. 11, the contrast quantification means 1 according to the third embodiment includes a blur image creation means 21 for creating blur image data Sus of the image data S0, and the blur image data Sus to a value of 0-15. 16-value conversion means 22 for 16-value conversion to obtain 16-value conversion blur image data Sus16, and a position for detecting a pixel position having a value of 15 as the maximum pixel value in the image represented by the 16-value conversion blur image data Sus16 As shown in FIG. 12, the distance between the pixel position detected by the detecting means 23 and the position detecting means 23 and the image center O is obtained, and the standard deviation σ of this distance is calculated, and this is used as the contrast feeling C0. And an arithmetic means 24. The contrast feeling C0 calculated in this way quantifies and represents the perceived state of how bright areas are distributed on the image when a human observes the image.

  Here, when the standard deviation σ calculated by the computing unit 24 is relatively small, an image in which bright areas are concentrated near the center of the image (for example, a strobe image obtained by shooting using a strobe). ), Image processing suitable for this is performed in the processing means 2. In the third embodiment, the processing unit 2 determines whether or not the contrast C0 obtained by the contrast quantification unit 1, that is, the standard deviation σ is smaller than a predetermined threshold Th5. If <Th5, it is determined that the image is a strobe image in which bright areas are concentrated near the center of the image, and image processing suitable for this is performed on the image data S0. When σ ≧ Th5, image processing suitable as a standard image is performed on the image data S0.

  Here, in the strobe image, since the subject is irradiated with strong light, the contrast is high, and the subject is white. Therefore, when it is determined that σ <Th5, the processing unit 2 extracts a bright area from the image represented by the image data S0, and the image data S0 in this area is, for example, shown in FIG. A gradation conversion process is performed using the LUT 5 so as to suppress the contrast. Thereby, it is possible to obtain processed image data S1 representing a non-jumping image while suppressing the contrast of the bright region.

  In the third embodiment, the distance between the pixel position detected by the position detector 23 and the center O of the image is obtained, and the standard deviation σ of this distance is used as the contrast feeling C0. As shown, an area A1 having a predetermined size is set near the center of the image, the number of pixels having a value of 15 in the area A1 is counted, and the number of pixels may be used as the contrast feeling C0. In this case, the processing means 2 determines whether or not the contrast C0, that is, the number of pixels having a value of 15 in the area A1 is greater than a predetermined threshold value Th6, and if it is determined that C0> Th6. Is assumed to be an image in which bright areas are concentrated near the center of the image, and gradation conversion processing is performed in the area A1 so as to suppress contrast. When it is determined that C0 ≦ Th6, it is assumed that the image is a standard image, and image processing suitable for this is performed. Thereby, similarly to the above, it is possible to suppress the contrast of the bright region portion and obtain processed image data S1 representing an image without a jump.

  In the first embodiment, the blurred image data Sus of the image data S0 is created, the histogram of the blurred image data Sus is obtained, and the contrast feeling C0 is quantified based on the histogram. S0 may be converted into a multi-resolution space for each of a plurality of frequency bands, a histogram of resolution data for each frequency band may be created, and the contrast C0 may be quantified based on the histogram. Hereinafter, this will be described as a fourth embodiment.

  FIG. 14 is a schematic block diagram showing a specific configuration of the fourth embodiment of the contrast quantification means 1. As shown in FIG. 14, the contrast quantification means 1 according to the fourth embodiment converts the image data S0 into a multi-resolution space by a wavelet transform, a Laplacian pyramid method, etc. Multi-resolution conversion means 31 for obtaining high-frequency band multi-resolution image data (hereinafter referred to as resolution data) RL, RM, and RH, and a region where the pixel value of the low-frequency band resolution data RL is a predetermined threshold Th7 or more The region extraction unit 32 that extracts the region M1, the histogram generation unit 33 that generates the histograms HM and HH of the region corresponding to the bright region M1 for the resolution data RM and RH in the middle and high frequency band, and the image data S0. Quantifying means 34 for quantifying the contrast C0 of the image to be obtained.

  In the contrast sensation quantifying means 1 according to the fourth embodiment, the contrast sensation C0 is quantified and obtained as follows. First, the image data S0 is converted into a multi-resolution space by the multi-resolution conversion means 31, and resolution data RL, RM, and RH in the low, medium, and high frequency bands are obtained. In the resolution data of each frequency band, the low resolution resolution data RL includes information on light and dark, but the medium and high resolution data RM and RH represent only frequency components. FIG. 15 is a diagram schematically showing an image represented by each resolution data. FIG. 15A shows the resolution data RL in the low frequency band, FIG. 15B shows the resolution data RM in the middle frequency band, and FIG. (C) shows an image represented by resolution data RH in a high frequency band.

  Subsequently, the resolution data RL in the low frequency band is input to the region extraction unit 32, where a region having a pixel value equal to or greater than a predetermined threshold Th <b> 7 is extracted as the bright region M <b> 1 and input to the histogram creation unit 33. On the other hand, the number n of pixels in the bright area M1 is input to the quantification means 34. Then, the histogram creation means 33 creates histograms HM and HH of the area corresponding to the bright area M1 for the resolution data RM and RH in the middle and high frequency bands. The histograms HM and HH are input to the quantification means 34.

  Here, the distribution width BL of the histogram HL of the resolution data RL in the low frequency band represents the distribution of pixel values as shown in FIG. 16A, and is similar to the histogram shown in FIG. 4 and FIG. The distribution widths BM and BH of the resolution data RM and RH in the middle and high frequency band represent the amplitude of the frequency centered at 0 as shown in FIGS. 16 (b) and 16 (c). Is.

  On the other hand, for example, the contrast of a face due to a nose or a dimple in the eyes and the shadow due to a building or a subject are composed of frequency components in a medium frequency band higher than the low frequency band. Therefore, an image having local contrast on a subject such as a face has a larger local shadow, and as a result, the amplitude of the resolution data RM in the middle frequency band increases. Further, for example, detailed structures such as tree branches and fineness of flowers, patterns and textures of human clothes, and boundaries (edges) between objects are composed of high frequency components. For this reason, the larger the contrast in the local region corresponding to the detailed structure, the more clearly these detailed structures can be seen, so the amplitude of the resolution data RH in the high frequency band increases.

  In the quantification means 34, the contrast feeling C0 is quantified based on the histograms HM and HH. First, the distribution width BM in the histogram HM of the resolution data RM in the middle frequency band is compared with a predetermined threshold Th8. When the distribution width BM of the histogram HM is larger than the predetermined threshold Th8 (BM> Th8), the standard image includes a relatively large amount of medium frequency band information, and is equal to or less than the predetermined threshold Th8 (BM ≦ Th8). In this case, it is determined that the image is a low contrast image that does not contain much information in the middle frequency band. If it is determined that the image is a standard image, the distribution width BH in the histogram HH of the resolution data RH in the high frequency band is compared with a predetermined threshold Th9, and the distribution width BH is larger than the predetermined threshold Th9 ( BH> Th9) may be determined to be a high-contrast image that contains a relatively large amount of high-frequency information, and a standard image that does not contain much high-frequency information if it is below a predetermined threshold Th9 (BH ≦ Th9). . On the other hand, the number n of pixels in the bright area M1 extracted by the area extraction unit 32 is compared with a predetermined threshold Th10. When the number n of pixels is smaller than the predetermined threshold Th10 (n <Th10), a low-contrast image. You may make it discriminate | determine that it is. In this case, when the number of pixels n is equal to or greater than the predetermined threshold Th10, the determination is performed as described above using the resolution data RM and RH in the middle and high frequency bands.

  When the type of contrast of the image is thus obtained, this type of contrast is output as a contrast feeling C0. In this case, the contrast feeling C0 is a signal having a value corresponding to the contrast type, for example, 1 for a low contrast image, 2 for a high contrast image, and 3 for a standard image.

  Then, the processing means 2 switches the gradation conversion LUT based on the contrast sensation C0 quantified by the contrast sensation quantification means 1 in the same manner as in the first embodiment, and the image data S0 is changed. Image processing S1 is obtained by performing image processing for converting the gradation.

In the fourth embodiment, as in the second embodiment, the luminance data L ^* or color data C ^* of the image data S0 is obtained, and the luminance data L ^* or color data C ^* is converted into a multi-resolution space. Then, resolution data for the luminance data L ^* or the color data C ^* may be obtained, and the contrast feeling C0 may be quantified from the resolution data. Hereinafter, this will be described as a fifth embodiment.

FIG. 17 is a schematic block diagram showing a specific configuration of the fifth embodiment of the contrast quantification means 1. As shown in FIG. 17, the contrast quantification means 1 according to the fifth embodiment is added to the multi-resolution conversion means 31, the region extraction means 32, the histogram creation means 33, and the quantification means 34 according to the fourth embodiment. The same conversion means 15 as that of the second embodiment is provided, and the multi-resolution conversion means 31 converts the luminance data L ^* or the color data C ^* into a multi-resolution space so that the low frequency band, medium frequency band, and high frequency band are converted. Luminance resolution data RLL, RML, RHL or color resolution data RLC, RMC, RHC are obtained.

  Here, when only the luminance resolution data RLL, RML, and RHL are obtained, the bright area M1 is extracted from the luminance resolution data RLL in the low frequency band by the area extracting unit 32 in the same manner as in the fourth embodiment. For the luminance resolution data RML and RHL in the frequency band, the luminance histograms HML and HHL of the area corresponding to the bright area M1 are created by the histogram creating means 33, and the quantifying means 34 based on the brightness histograms HML and HHL. C0 is quantified.

  On the other hand, when only the color resolution data RLC, RMC, and RHC are obtained, the area extraction unit 32 extracts an area that is equal to or higher than a predetermined threshold from the image represented by the color resolution data RLC in the low frequency band as the high saturation area M2. .

  Here, the color resolution data RMC in the middle frequency band represents, for example, the brightness of the face due to a nose or a dimple in the eye and the shadow due to a building or subject, like the resolution data RM in the middle frequency band of the image data S0. Similarly to the resolution data RH of the high frequency band of the image data S0, the color resolution data RHC of the high frequency band, for example, the fineness of tree branches and flowers, the pattern and texture of a person's clothes, and the boundary (edge) between objects ) And other detailed structures.

  Therefore, the histogram creation means 33 creates color histograms HMC and HHC of the medium and high frequency band color resolution data RMC and RHC for the high saturation region M2 extracted by the region extraction means 32, and the quantification means 34 describes the fourth color histogram HMC and HHC. As in the first embodiment, the contrast type is determined by comparing the amplitudes of the color histograms HMC and HHC with a predetermined threshold, and the determination result can be obtained as the contrast C0.

Further, in the fifth embodiment, when the contrast feeling C0 is obtained from the color data C ^* , when the contrast type obtained from the contrast feeling C0 is a low contrast image, the image represented by the image data S0. You may make it perform the saturation change process which raises the saturation of. Specifically, the saturation is improved by multiplying the saturation C ^* obtained by the above equation (8) by the enhancement coefficient αc. Note that the value of the enhancement coefficient αc is preferably about 1.2, but is not limited thereto.

  Furthermore, in the fifth embodiment, when the luminance resolution data RLL, RML, RHL and the color resolution data RLC, RMC, RHC are obtained, the bright area M1 is extracted based on the luminance resolution data RLL in the low frequency band, For the color resolution data RMC and RHC in the middle and high frequency bands, the color histograms HMC and HHC of the area corresponding to the bright area M1 may be created, and the contrast C0 may be quantified based on the color histograms HMC and HHC. Further, the high saturation region M2 is extracted based on the color resolution data RLC in the low frequency band, and the brightness histograms HML and HHL of the region corresponding to the high saturation region M2 are created for the brightness resolution data RML and RHL in the medium and high frequency band to create contrast. The feeling C0 may be determined by quantification.

In the first and third embodiments, the distribution width of the histogram of the image data S0, the luminance data L ^* and the color data C ^* , and the standard deviation of the image data S0 in the second embodiment are the contrast feeling C0. However, for one piece of image data S0, the image data S0, the luminance data L ^* , the color data C ^* , the distribution width and standard deviation of the histogram, and various information that can determine the contrast of the image data S0. As a feature value, the contrast C0 may be quantified by weighting and adding each feature value as shown in the following equation (9).

Where vn: feature quantity an: weighting coefficient
n: Number of feature quantities The weighting factor an may be obtained experimentally. That is, the contrast feeling C0 is quantified by variously changing the weighting coefficient, and a plurality of images subjected to different image processing according to the contrast feeling C0 are generated for visual evaluation, and an image that matches the visual contrast feeling. May be selected and the selected weighting factor an may be used in the calculation of equation (9).

  In the first to third embodiments, the contrast feeling C0 is obtained by quantification. However, as in the fourth and fifth embodiments, the image contrast type may be obtained as the contrast feeling C0. Good.

Furthermore, in the first to third embodiments, as shown in the following equation (10), the type of contrast of the image is determined according to the weighted addition result of the feature quantity vn, and this determination result is used as the contrast feeling. It may be C0.

Here, hn, sn, and ln are weighting factors that are calculated so that Ph, Ps, and Pl are indexes representing the probabilities of being a high-contrast image, a standard image, and a low-contrast image, respectively. For example, when Ph, Ps, and Pl are calculated as 10%, 30%, and 70% for a certain image and calculated as 80%, 40%, and 5% for another image, the former is a low contrast image, and the latter is a high contrast. It can be determined that it is an image. In this case, the contrast feeling C0 may be a signal having a value corresponding to the contrast type, such as 1 for a low contrast image, 2 for a high contrast image, and 3 for a standard image, and is represented by the contrast feeling C0. A gradation conversion LUT may be selected according to the contrast type and gradation conversion processing may be performed on the image data S0.

In the fourth embodiment, the number n of pixels in the bright area M1 obtained from the resolution data RL in the low frequency band and the distribution widths BM and BH of the histograms HM and HH of the resolution data RM and RH in the middle and high frequency bands. The sense of contrast C0 is quantified based on the above, but the resolution data RL, RM, and RH of each frequency band are obtained in the number of pixels in the bright area, the distribution width of the histogram, or in the third embodiment. A standard deviation may be obtained as a feature amount for each frequency band, and the feature amount may be weighted and added by the above equation (9) to quantify the contrast C0 for each frequency band. Furthermore, as shown in the following formula (11), the contrast feeling C0 may be quantified by further weighted addition of the contrast feeling obtained individually for each frequency band. The feature amount includes the number n of pixels in the bright area M1 obtained from the low-frequency band luminance resolution data RLL and the luminance histograms HML and HHL of the medium-high frequency band luminance resolution data RML and RHL in the fifth embodiment. And / or the distribution widths of the color histograms HMC and HHC of the color resolution data RMC and RHC in the medium and high frequency band in the high saturation region M2 obtained from the color resolution data RLC in the low frequency band.

However, li, mj, hk: feature quantities for each frequency band αi, βj, γk: weighting coefficients corresponding to an in the above formula (9) L, M, H: weighting coefficients i, j, for each frequency band k: Number of feature values in each frequency band The weighting factors αi, βj, γk, L, M, and H may be obtained experimentally in the same manner as the weighting factor an in Equation (9). Further, if L = 1, M = 0, and H = 0 in equation (11), this is substantially equivalent to equation (9).

  Furthermore, in this case, the images are classified into types such as a high-contrast image, a standard image, and a low-contrast image according to the value of the contrast feeling C0 calculated in the above formulas (9) and (11). May be set as a contrast feeling C0. In this case, the contrast C0 may be a numerical value representing a high-contrast image, a standard image, and a low-contrast image.

  Further, in the fourth and fifth embodiments, the probability of the contrast type is obtained as in the above equation (10), the contrast type of the image is discriminated based on this probability, and the discrimination result is obtained as the contrast feeling C0. May be.

In each of the above embodiments, the processing means 2 performs the gradation conversion process and / or the saturation change process by switching the gradation conversion LUT according to the contrast C0 for the image data S0. The image processing is not limited to this. For example, the contrast feeling C0 is compared with a predetermined threshold value Th11, and when C0 <Th11, the frequency component in the hatched portion is enhanced by the following formula (12) as shown in FIG. Such frequency processing may be performed on the image data S0. In FIG. 18, the Fx axis and the Fy axis represent frequencies on the Fourier plane, and the hatched portion corresponds to a high frequency component in the image data S0.

However, F (x, y): Image data S0
F ′ (x, y): processed image data S1
d (C0): a function that determines the degree of blur β: enhancement coefficient Further, when the image data S0 represents an image including a human face, a face region corresponding to the face is extracted from the image data S0, The processed image data S1 may be obtained by performing an AE process (automatic exposure control process) so as to change the brightness of the face area in accordance with the density of the area and the quantified contrast C0. Hereinafter, processing means for performing this AE process will be described. FIG. 19 is a schematic block diagram showing the configuration of the processing means 2 that performs AE processing. As shown in FIG. 19, the processing means 2 obtains a face extracting means 41 for extracting a human face area from an image represented by the image data S0, and a face area density tface extracted by the face extracting means 41. Density calculating means 42, and AE processing means 43 for performing AE processing on the image data S0 based on the contrast C0 and the face area density tface to obtain processed image data S1.

  As a face region extraction method in the face extraction means 41, as described in, for example, Japanese Patent Laid-Open No. 6-67320, an image is extracted based on the hue and saturation value distribution of the image represented by the image data S0. Extract the face candidate area by dividing the area and then detect and extract the face area from the shape of the neighboring area located in the vicinity of the face candidate area, or simply find the ellipse that circumscribes the extracted face candidate area A method of setting a region surrounded by the ellipse as a face region can be employed. Further, for example, a method of extracting a face region by a neural network described in Japanese Patent Application Laid-Open Nos. 5-274438 and 5-307605 may be used.

  In the density calculation means 42, the average value of the pixel values in the face area extracted by the face extraction means 41 is calculated as the face area density tface.

  When the AE process is performed, the contrast C0 input to the processing unit 2 represents the distribution width of the histogram Hus as quantified in the first embodiment.

  Then, the face extraction means 41 extracts a face area from the image represented by the image data S0, and the density calculation means 42 calculates the density tface of the face area. The AE processing unit 43 performs AE processing on the image data S0 based on the contrast C0 and the face area density tface. In the AE processing unit 43, after the AE process is once performed on the entire image data S0, the AE process is further performed only on the face area in order to further optimize the face density. FIG. 20 is a diagram for explaining the AE process. In the AE processing means 43, the contrast feeling C0 is first compared with predetermined threshold values Th12 and Th14 (Th12 <Th14). If C0 <Th12, it is determined that the image is a low contrast image, and the face area density tface is further compared with a predetermined threshold Th13. If tface <Th13, it is assumed that the face area is too dark, and the AE process is performed on the image data S0 so that the face area density tface matches Th13 as shown in FIG. Processed image data S1 is obtained.

  If C0 <Th12 and tface ≧ Th13, the brightness of the face area is appropriate and the AE process is not performed.

  In the case of C0 ≧ Th12 and tface <Th13, the AE process is performed on the image data S0 so that the face area density tface is (Th13 + tface) / 2, assuming that the contrast C0 is appropriate but the face area is too dark. To obtain processed image data S1.

  When C0 ≧ Th12 and tface ≧ Th13, the brightness of the face area is appropriate and the AE process is not performed.

  On the other hand, if C0> Th14, it is determined that the image is a high contrast image, and the face area density tface is further compared with a predetermined threshold Th15. If tface> Th15, it is determined that the face area is too bright, and the processed image data S1 is obtained by performing AE processing so that the face area density tface matches Th15 as shown in FIG. obtain. In this case as well, it is preferable to perform the AE process only on the face area.

  If C0> Th14 and tface ≦ Th15, the brightness of the face area is appropriate and the AE process is not performed.

  When C0 ≦ Th14 and tface> Th15, it is assumed that the contrast C0 is appropriate but the face area is too bright, and the AE process is performed on the image data S0 so that the face area density tface is (Th15 + tface) / 2. To obtain processed image data S1.

  When C0 ≦ Th14 and tface ≦ Th15, the brightness of the face area is appropriate and the AE process is not performed.

  In the third embodiment, assuming that σ <Th5, the AE process is performed so that the density tface of the face area becomes small, assuming that the face area is a white strobe image. Good.

  In each of the above embodiments, the processing means 2 performs gradation conversion processing, saturation change processing, frequency enhancement processing, or AE processing. For example, the gradation conversion processing and AE processing are the same. These processes may be performed in combination. Further, the content of the image processing is not limited to the gradation conversion processing, the saturation change processing, the frequency enhancement processing, and the AE processing, and various other image processing can be performed.

Next, another embodiment according to the present invention will be described. FIG. 21 is a schematic block diagram showing the configuration of an image processing apparatus according to another embodiment of the present invention. As shown in FIG. 21, an image processing apparatus according to another embodiment of the present invention includes a conversion unit 51 that converts image data S0 into luminance data L ^* and color data C ^* , similar to the conversion unit 15 in the second embodiment. And multi-resolution conversion of luminance data L ^* and color data C ^* to obtain luminance resolution data RLL, RML, RHL and color resolution data RLC, RMC, RHC in the low frequency band, medium frequency band and high frequency band Based on the conversion means 31, the histogram creation means 53 for creating the luminance histogram HLL of the luminance resolution data RLL in the low frequency band and the color histogram HLC of the color resolution data RLC in the low frequency band, and the luminance histogram HLL and the color histogram HLC, Processing pattern for setting image processing pattern J to be applied to image data S0 Setting means 54, processing means 55 for performing image processing on the image data S0 by the set processing pattern J to obtain processed image data S1, and output means for outputting the processed image data S1 as a visible image 56.

When the color data C ^* is obtained as saturation, a one-dimensional color histogram HLC is created, and when the color data is obtained as a ^* , b ^* , a two-dimensional color as shown in FIG. A histogram HLC is created.

In the processing pattern setting means 54, the value of the enhancement coefficient αc for improving the gradation conversion LUT and the saturation C ^* shown in FIG. First, a feature amount representing a histogram distribution is obtained from the luminance histogram HLL and the color histogram HLC. The distribution width of the luminance histogram HLL is obtained as the feature amount P1. If the color histogram HLC is one-dimensional, the distribution width is obtained as the feature amount P2, and if the color histogram HLC is two-dimensional, the distribution area is obtained as the feature amount P2. Then, the feature amount P2 is compared with a predetermined threshold Th16, and if P2 ≧ Th16, for example, LUT3 of the tone conversion LUT shown in FIG. 6 is selected as the LUT for tone conversion. Further, the enhancement coefficient αc = 1.0 is set.

  On the other hand, when P2 <Th16, the feature amount P1 obtained from the luminance histogram HLL is compared with a predetermined threshold Th17, and when P1 ≧ Th17, the LUT 5 is selected as a high contrast image. And the enhancement coefficient αc = 1.0 is set. If P1 <Th17, LUT1 is selected as a low contrast image, and the enhancement coefficient αc = 1.2 is set.

  The processing unit 55 performs image processing on the image data S0 according to the processing pattern J set in this way.

Next, the operation of another embodiment will be described. FIG. 22 is a flowchart showing the operation of another embodiment. First, the conversion means 51 converts the image data S0 into luminance data L ^* and color data C ^* (step S11), and the multiresolution conversion means 52 converts the luminance data L ^* and color data C ^* into multiresolution. Luminance resolution data RLL, RML, RHL and color resolution data RLC, RMC, RHC in the low frequency band, medium frequency band, and high frequency band are obtained (step S12). In the histogram creation means 53, the brightness histogram HLL is created from the brightness resolution data RLL in the low frequency band, and the color histogram HLC is created from the color resolution data RLC in the low frequency band (step S13).

  Then, the feature amounts P1 and P2 of the luminance histogram HLL and the color histogram HLC are calculated (step S14), and it is determined whether P2 ≧ Th16 (step S15). When step S15 is affirmed, LUT3 is selected, the enhancement coefficient αc is set to 1.0 (step S16), and the processing pattern J is set. If step S15 is negative, it is determined whether P1 ≧ Th17 (step S17). If step S17 is positive, LUT5 is selected and the enhancement coefficient αc is set to 1.0. In step S18, the processing pattern J is set. If step S17 is negative, LUT1 is selected, the enhancement coefficient αc is set to 1.2 (step S19), and the processing pattern J is set. Then, image processing is performed on the image data S0 by the set processing pattern J (step S20), and processed image data S1 is obtained. The obtained processed image data S1 is output as a visible image by the output means 56 (step S21).

In the other embodiment, the luminance data L ^* and the color data C ^* are subjected to multi-resolution conversion, and the histogram HLL of the luminance resolution data RLL in the low frequency band and the histogram HLC of the color resolution data RLC in the low frequency band are generated. However, as shown in FIG. 23, instead of the multi-resolution conversion means 52, luminance blurred image data Lus and color blurred image data Cus, which are blurred image data of the luminance data L ^* and the color data C ^* , are created. The blur image creating means 57 may be provided to create the brightness histogram HLus from the brightness blur image data Lus and the color histogram HCus from the color blur image data Cus.

1 is a schematic block diagram showing the configuration of an image processing apparatus according to an embodiment of the present invention. Schematic block diagram showing the configuration of the first embodiment of the contrast quantification means Diagram showing the creation state of blurred image data The figure which shows the example of the histogram of a blur image data The figure which shows the example of the histogram of a blur image data The figure which shows gradation conversion LUT The flowchart which shows the process performed in 1st Embodiment. The figure which shows the histogram after bokeh image data is 16-valued Schematic block diagram showing the configuration of the second embodiment of the contrast quantification means Figure showing a two-dimensional histogram Schematic block diagram showing the configuration of the third embodiment of the contrast quantification means The figure for demonstrating the process performed in 3rd Embodiment The figure for demonstrating the process performed in the modification of 3rd Embodiment. Schematic block diagram showing the configuration of the fourth embodiment of the contrast quantification means The figure which shows the image represented by the resolution data of each frequency band The figure which shows the histogram of the resolution data of each frequency band Schematic block diagram showing the configuration of the fifth embodiment of the contrast quantification means Diagram for explaining frequency enhancement processing Schematic block diagram showing the configuration of processing means for performing AE processing Diagram for explaining AE processing The schematic block diagram which shows the structure of the image processing apparatus by other embodiment of this invention. The flowchart which shows the process performed in other embodiment The schematic block diagram which shows the structure of the image processing apparatus by further another embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Contrast feeling quantification means 2,55 Processing means 3,56 Output means 11,21 Blurred image creation means 12,53 Histogram creation means 13,34 Quantification means 15,51 Conversion means 22 16-value conversion means 23 Position detection means 24 Calculation means 31, 52 Multi-resolution conversion means 32 Region extraction means 33 Histogram creation means 41 Face extraction means 42 Density calculation means 43 AE processing means 54 Processing pattern setting means

Claims (3)

Conversion means calculates color data representing color information of an image represented by the image data from the image data;
A blur image creating means creates blur image data of the color data,
A histogram creating means creates a histogram representing a two-dimensional frequency distribution of the blurred image data;
The contrast feeling calculating means calculates the distribution area of the histogram as a contrast feeling,
An image processing method, wherein processing means performs image processing for changing luminance information of the image on the image data based on the sense of contrast. Conversion means for calculating color data representing color information of an image represented by the image data from the image data;
A blurred image creating means for creating blurred image data of the color data;
A histogram creating means for creating a histogram representing a two-dimensional frequency distribution of the blurred image data;
Contrast feeling calculating means for calculating the distribution area of the histogram as a contrast feeling;
An image processing apparatus comprising: processing means for performing image processing for changing luminance information of the image on the image data based on the sense of contrast. Conversion means for calculating color data representing color information of an image represented by the image data from the image data;
A blurred image creating means for creating blurred image data of the color data;
A histogram creating means for creating a histogram representing a two-dimensional frequency distribution of the blurred image data;
Contrast feeling calculating means for calculating the distribution area of the histogram as a contrast feeling;
A program that functions as processing means for performing image processing for changing luminance information of the image on the image data based on the sense of contrast.

Images