JP2007144641A - Image processing technique - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processing technique which can properly perform the judgement for a similarity between images by taking a difference in the brightness between the images by the presence/absence of a flash under consideration. <P>SOLUTION: A printer 10 includes an image reading section 111, an image specifying section 112, a description amount extracting section 113, a similarity/non-similarity judging section 116, an image selecting section 117, and a printing control section 118. In this case, the image reading section 111 reads a plurality of images from a storage medium 810. The image specifying section 112 specifies two images from the plurality of images. The description amount extracting section 113 extracts description amounts indicating the descriptions of colors which constitute the images. The similarity/non-similarity judging section 116 judges the similarity of the two images based on the description amounts. The image selecting section 117 selects the images from among the plurality of images based on the judged similarity. The printing control section 118 performs a control for printing the selected images. The similarity/non-similarity judging section 116 judges the similarity by handling an approximate value which is approximate to the description amount within a range of estimated values, for which the description amount is estimated to vary depending on the absence of a flash, when it is judged that the presence/absence of the flash is related. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image processing technique for determining similarity between images.

  As an image processing apparatus that handles a plurality of images, for example, a personal computer or a digital still camera, a printing apparatus (direct printer) capable of directly printing an image stored in a storage medium, or a storage medium There are image display devices (photo viewers) that display images. Conventionally, in these image processing apparatuses, in order to improve convenience of image management, printing, viewing, and the like, an image processing technique for determining similarity between images has been proposed. Patent Document 1 below discloses a conventional image processing technique for determining the similarity between images based on the color histogram of each image.

JP 2000-137722 A

However, in the conventional image processing technology, the color histograms of the images are compared as they are to determine the similarity between the images. Therefore, if the brightness of the images varies greatly depending on the presence or absence of a flash at the time of shooting, the images are determined to be similar. In some cases, power images should be dissimilar. For example, in the case of images of the same scene taken continuously by changing shooting conditions such as exposure change or presence / absence of flash, there is a demand that such images of the same scene should be judged to be similar. Nevertheless, with the conventional image processing technique, if the brightness between images differs greatly depending on the presence or absence of a flash, even images in the same scene may be determined to be dissimilar.

  In view of the above-described problems, an object of the present invention is to provide an image processing technique capable of appropriately determining similarity between images in consideration of a difference in brightness between images depending on the presence or absence of a flash. .

  In order to solve the above-described problems, a printing apparatus of the present invention is a printing apparatus that prints some of a plurality of images stored in a storage medium, an image reading unit that reads the plurality of images from the storage medium, and a reading An image specifying unit that specifies two images from a plurality of images, a feature amount extracting unit that extracts a feature amount indicating a color feature constituting the image from each of the two specified images, and A similarity determination unit that determines the similarity between two images based on the two feature amounts, an image selection unit that selects an image from a plurality of images based on the determined similarity, and the selected image A similarity control unit that controls whether one of the two specified images is shot with a flash and the other is shot without a flash. To determine whether When determining that there is a flash presence / absence relationship with the flash determination unit, an approximate value that approximates the other of the feature amounts within a range of values in which one of the two feature amounts is assumed to change depending on the presence / absence of flash. , And a feature quantity replacement unit that determines similarity by treating the feature quantity as one of the feature quantities. This can reduce the influence of the presence or absence of flash on the determination of the similarity between images, so that the similarity between images can be determined appropriately. As a result, it is possible to appropriately execute image printing according to the similarity.

  The above-mentioned printing apparatus can also take the following forms. For example, one of the feature amounts may be a feature amount extracted from an image shot with a flash. As a result, the assumed error of the value that the feature value with flash is assumed to change with no flash is smaller than the assumed error with the value with which the feature value without flash is assumed to change with flash. The similarity can be judged more appropriately.

  Further, the assumed value may be a value indicating a feature of a color that is the same color as that indicated by one of the feature amounts and has a different brightness. Thereby, an approximate value handled as one of the feature values can be easily calculated.

  The storage medium stores flash information indicating the presence or absence of a flash when an image is captured for each image, and the flash determination unit determines whether or not there is a flash presence or absence based on the flash information. You may judge. As a result, it is possible to accurately determine whether or not there is a flash presence / absence relationship. The flash information may be Exif standard information.

  The feature amount extraction unit includes a block unit that divides the two specified images into a plurality of blocks, and an RGB extraction unit that extracts an RGB value indicating a representative color for each of the plurality of blocks as a feature amount. The similarity determination unit may determine the similarity between two images based on the distance in the RGB Euclidean space for the two RGB values respectively extracted from the corresponding blocks between the two images. Accordingly, it is possible to determine the similarity between images based on the similarity of the arrangement of colors constituting each image.

  The image selection unit includes an image classification unit that classifies a plurality of images into groups for each similar image based on the determined similarity between the two images, and is included in the group for each classified group. An image selection unit that selects at least one of the images as the selected image may be provided. As a result, it is possible to prevent similar images from being printed repeatedly among a plurality of images stored in the storage medium.

  Further, in a form in which a plurality of images are grouped, the plurality of images stored in the storage medium are images arranged in a series in the order of shooting, and the two specified images are adjacent in the order of shooting. It may be images. As a result, it is not necessary to determine the similarity for all the combinations of images, and the images can be grouped by determining the similarity between adjacent images in the order of shooting. Processing efficiency can be improved.

  Further, the image selection unit is configured to search an image similar to one of the identified images from a plurality of images based on the determined similarity between the two images, and an image search unit You may provide the image selection part which selects at least one as the selected image. Thus, an image similar to a specific image among a plurality of images stored in the storage medium can be printed.

  Further, an evaluation value extraction unit that extracts an evaluation value indicating the brightness of the image from each of the two specified images, and at least one of the two feature amounts extracted based on the extracted evaluation value May be provided with a feature amount correction unit that corrects the two images in the direction in which the brightness approaches. This can reduce the influence of differences in the brightness of the entire image between images on the determination of similarity between images, so that the determination of similarity between images can be performed more appropriately.

  Further, in the embodiment in which the feature amount is corrected based on the evaluation value, the shooting condition when the image is shot is stored for each image in the storage medium, and the evaluation value extraction unit evaluates based on the shooting condition. A value may be extracted. Thereby, the influence of the difference in the brightness of the entire image between the images on the determination of the similarity between the images can be reduced according to the photographing condition of each image.

  In the form in which the evaluation value is extracted based on the shooting condition, the shooting condition is an aperture value, shutter speed, and ISO sensitivity, and the evaluation value indicates the luminance of the subject based on the aperture value, shutter speed, and ISO sensitivity. It may be a value. Thereby, the influence of the difference in the brightness of the entire image between the images on the determination of the similarity between the images can be reduced based on the luminance of the subject captured in each image. In the form in which the evaluation value is extracted based on the shooting condition, the shooting condition may be Exif standard information.

  In the form in which the feature amount is corrected based on the evaluation value, the evaluation value extraction unit may extract the evaluation value based on the luminance of the image itself. As a result, it is possible to reduce the influence of the difference in the brightness of the entire image between the images on the determination of the similarity between the images based on the luminance of each image itself. In the form in which the evaluation value is extracted based on the luminance of the image itself, the evaluation value may be a degree of luminance based on a cumulative histogram of luminance for the image.

  Further, in the embodiment in which the feature amount is corrected based on the evaluation value, the feature amount correction unit converts at least one of the two extracted feature amounts based on the extracted evaluation value in the direction in which the image becomes brighter. You may correct | amend in the direction which the brightness of an image approaches. As a result, correction is always performed in the direction in which the image is brightened, that is, in the direction in which the feature amount increases, so an image clipped in advance to the maximum value that the feature amount can take and an image not clipped in advance are equivalent. Can be compared.

  Further, in the form in which the feature amount is corrected in the direction in which the image becomes brighter, the feature amount correction unit sets the corrected feature amount to the maximum value when the corrected feature amount exceeds the maximum value in the direction in which the image becomes brighter. A clipping unit for correction may be provided. As a result, an image clipped in advance to the maximum value that can be taken by the feature value and an image in which the feature value exceeds the maximum value due to the correction by the evaluation value can be compared with each other.

  Further, in the embodiment in which the feature amount is corrected based on the evaluation value, the feature amount correction unit converts one of the two extracted feature amounts based on the extracted evaluation value, and one brightness of the two images. You may correct | amend in the direction close | similar to the other brightness. Thereby, it is possible to improve the processing efficiency in correcting the feature value by the evaluation value as compared with the case of correcting both of the two feature values.

  Further, in the embodiment in which the feature amount is corrected based on the evaluation value, the feature amount extraction unit includes a block unit that divides each of the two specified images into a plurality of blocks, and an RGB value that represents a representative color for each of the plurality of blocks. And an RGB extraction unit that extracts the feature value as a feature amount. The feature amount correction unit includes a gamma correction unit that gamma-corrects the RGB value extracted as the feature amount, and the gamma-corrected feature amount as an extracted evaluation value. Based on this, a conversion correction unit that performs level correction in a direction in which the brightness of two images approaches and a reverse gamma correction unit that performs reverse gamma correction on the level-corrected feature amount may be provided. Thereby, the correction of the feature amount based on the evaluation value can be performed more strictly.

  Further, in the embodiment in which the feature amount is corrected based on the evaluation value, the feature amount correction unit prohibits correction of the feature amount based on the evaluation value when the difference between the two extracted evaluation values is less than the reference value. A correction prohibition unit may be provided. As a result, when the difference in brightness between images does not affect the determination of similarity, it is not necessary to perform feature amount correction based on the evaluation value, thereby improving the processing efficiency in determining the similarity between images. Can be made.

  In order to solve the above-described problem, an image processing apparatus according to the present invention is an image processing apparatus that processes a plurality of images stored in a storage medium, and an image reading unit that reads the plurality of images from the storage medium; An image specifying unit for specifying two images from the plurality of read images, a feature amount extracting unit for extracting a feature amount indicating a color feature constituting the image from each of the two specified images, A similarity determination unit that determines the similarity between the two images based on the two extracted feature quantities. The similarity determination unit captures one of the two specified images with a flash and the other flashes. A flash determination unit that determines whether there is a flash presence / absence relationship, and when it is determined that there is a flash presence / absence, one of the two feature quantities varies depending on the presence / absence of the flash. Then an approximate value which approximates to the other feature amounts within the range of assumed values, by treating as one of the characteristic amount, characterized by comprising a feature amount replacement section for determining similarity. This can reduce the influence of the presence or absence of flash on the determination of the similarity between images, so that the similarity between images can be determined appropriately.

  The above-mentioned image processing apparatus can also take the following forms. For example, the image processing apparatus described above may further include an image classification unit that classifies a plurality of images into groups for each similar image based on the determined similarity between the two images. As a result, the classification accuracy of groups for each similar image can be improved.

  Further, the above-described image processing apparatus further includes an image search unit that searches an image similar to one of the specified images from a plurality of images based on the determined similarity between the two images. Also good. Thereby, the search accuracy of similar images can be improved.

  Further, the form of the present invention is not limited to the form of a printing apparatus or an image processing apparatus. For example, an image processing method for handling a plurality of images stored in a storage medium and a function for processing a plurality of images in a computer. It can also be applied to a program for realizing it. Further, the present invention is not limited to the above-described embodiments, and it is needless to say that the present invention can be implemented in various forms without departing from the spirit of the present invention.

  In order to further clarify the configuration and operation of the present invention described above, an image processing technique to which the present invention is applied will be described below.

A. Example:
A-1. Configuration of printing device:
FIG. 1 is a perspective view showing an external configuration of the printing apparatus 10. FIG. 2 is an explanatory diagram mainly showing the internal configuration of the printing apparatus 10. A printing apparatus 10 shown in FIGS. 1 and 2 is an apparatus having various functions such as a color printer, a monochrome printer, a scanner, and a copying machine, that is, a so-called multifunction machine. In addition to these functions, the printing apparatus 10 has a direct print function for converting an image stored in a storage medium connected to the printing apparatus 10 into print data and printing it directly.

  The printing apparatus 10 includes a main control unit 110 that controls each unit of the printing apparatus 10, a printing unit 120 that performs printing on a printing medium 900 such as plain paper, photographic paper, and a postcard, and reads an image on a paper surface into digital data. A scanner unit 130 for conversion (scanning), a card slot 140 capable of reading and writing data to a storage medium 810 such as a memory card incorporating a flash memory, and a storage medium such as a personal computer, a digital still camera, and a digital video camera A communication connector 150 capable of exchanging data with an external device 820 having a built-in device, a display unit 160 that displays characters and images, and an input button unit 170 that receives an instruction input from a user of the printing apparatus 10.

  The main control unit 110 of the printing apparatus 10 includes an image reading unit 111 that reads a plurality of image files 710 from a storage medium 810 connected to the card slot 140 and an external device 820 connected to the communication connector 150, and a plurality of image files 710. The image specifying unit 112 that specifies two images PI for judging similarity from among a plurality of images PI reproduced by the above, and the feature amount extraction that extracts the feature amount F indicating the feature of the color constituting the image PI. Unit 113, an evaluation value extraction unit 114 that extracts an evaluation value L indicating the brightness of the image PI, a feature amount correction unit 115 that corrects the feature amount F based on the evaluation value L, and two based on the feature amount F A similarity determination unit 116 that determines the similarity of two images PI, an image selection unit 117 that selects an image PI from a plurality of images PI based on the determined similarity, and a selection And a print control unit 118 controls to print the image PI.

  In the present embodiment, the main control unit 110 includes a central processing unit (hereinafter referred to as CPU), a read only memory (hereinafter referred to as ROM), a random access memory (hereinafter referred to as ROM), and a RAM. ASIC (Application Specific Integrated Circuits) provided with hardware such as. Details of the operation of the main controller 110 will be described later.

  As shown in FIG. 2, the printing unit 120 of the printing apparatus 10 includes a print head unit 210 that discharges printing ink onto a print medium 900, a carriage 200 that mounts the print head unit 210, and drives the carriage 200 in the main scanning direction. A carriage drive unit 240 for transporting the print medium 900 and a print medium transport unit 250 for transporting the print medium 900. The print head unit 210 of the printing unit 120 includes a total of six ink ejection heads 211 to 216 for each color ink of black, cyan, light cyan, magenta, light magenta, and yellow. Each of the ink ejection heads 211 to 216 ejects ink onto the print medium 900 by adjusting the voltage of a piezo element (not shown). Each of the ink ejection heads 211 to 216, the carriage drive unit 240, and the print medium transport unit 250 of the printing unit 120 implements printing on the print medium 900 by cooperating based on an instruction from the main control unit 110.

  FIG. 3 is an explanatory diagram schematically showing the data structure of the image file 710. Each of the image files 710 includes shooting condition data 7110 indicating shooting conditions when the image PI played back by the image file 710 is shot, and image data 7120 indicating the image PI played back by the image file 710. Have. The shooting condition data 7110 of the image file 710 includes, as information indicating shooting conditions, a shooting date and time 7112 when the image PI was shot, an aperture value (F value) 7114 when the image PI was shot, and the image PI. Has a shutter speed 7116 when the image PI is photographed, ISO sensitivity 7118 when the image PI is photographed, and flash information 7119 indicating whether or not the image PI is photographed using a flash. In this embodiment, the image file 710 is an Exif standard JPEG file in which an image PI photographed by a digital still camera is recorded.

A-2. Printer behavior:
FIG. 4 is a flowchart showing image printing processing executed by the main control unit 110 of the printing apparatus 10. The main control unit 110 of the printing apparatus 10 includes a storage medium 810 and an external device 820 attached to the card slot 140 and the communication connector 150 by the user, and a request for executing the direct print function is input by the user via the input button unit 170. 4 starts the image printing process shown in FIG.

  When starting the image printing process of FIG. 4, the main control unit 110 of the printing apparatus 10 reads a plurality of image files 710 from the storage medium 810 and the external device 820 attached to the card slot 140 and the communication connector 150 (step S100). ). Thereafter, the main control unit 110 classifies the plurality of images PI reproduced from the plurality of read image files 710 into groups for each similar image (step S150). It identifies in order (steps S110 and S158).

  FIG. 5 is an explanatory diagram showing how two images PI are specified in the image printing process of FIG. 4 (step S110). In the present embodiment, the main control unit 110 specifies a plurality of images PI that are arranged in a sequence in the order in which the images were captured, by shifting the images PI that are adjacent in the order in which the images were captured one by one from the top image PI. Step S110). That is, the second image PI captured later among the two specified images PI is the first image captured first among the two images PI when the two images PI are specified next. Specified as an image PI.

  Returning to the description of FIG. 4, after the two images PI are identified (step S <b> 110), the main control unit 110 extracts an evaluation value L indicating the brightness of the image PI from each of the identified two images PI. An evaluation value extraction process is executed (step S120).

  FIG. 6 is a flowchart showing the evaluation value extraction process (step S120). When the evaluation value extraction process of FIG. 6 is started, the main control unit 110 reads out the aperture value 7114, the shutter speed 7116, and the ISO sensitivity 7118 as the shooting conditions from each of the image files 710 in which the two images PI are recorded (step) S1202). Thereafter, the main control unit 110 calculates the luminance of the subject based on the read aperture value 7114, shutter speed 7116, and ISO sensitivity 7118 as the evaluation value L (step S1204). In this embodiment, the main control unit 110 converts each of the aperture value 7114, the shutter speed 7116, and the ISO sensitivity 7118 into the APEX unit system using the following formula 1, and then uses the following formula 2 to An evaluation value L is calculated from the converted value.

  Here, the evaluation value L calculated by Expression 2 means that when the value is reduced by “1”, the subject of the image PI is photographed with double brightness. In the following description, of the two extracted evaluation values L, the smaller one is expressed as an evaluation value Lt, and the larger one is expressed as an evaluation value Lc. Among them, the brighter side, that is, the one from which the evaluation value Lt is extracted is denoted as an image TI, and the darker side, that is, the one from which the evaluation value Lc is extracted is denoted as an image CI.

  Returning to the description of FIG. 4, after the evaluation values Lt and Lc are extracted (step S120), the main control unit 110 determines the feature amount F indicating the color feature constituting the image PI as the two identified images TI. , CI is extracted from each of the CIs (step S130). In the following description, of the two extracted feature values F, the one extracted from the image TI is expressed as a feature value Ft, and the one extracted from the image CI is expressed as a feature value Fc.

  FIG. 7 is a flowchart showing the feature amount extraction processing (step S130). When starting the feature amount extraction processing of FIG. 7, the main control unit 110 divides each of the two specified images TI and CI into a plurality of blocks (step S1302), and becomes a key in each of the divided blocks. A representative color that is a color is extracted as feature amounts Ft and Fc (step S1304).

  FIG. 8 is an explanatory diagram showing the feature amount extraction process (step S130) of FIG. The left side of FIG. 8 shows how the feature quantity Fc is extracted from the image CI, and the right side of FIG. 8 shows how the feature quantity Ft is extracted from the image TI. In this embodiment, the main control unit 110 divides each of the two images TI and CI into blocks of 3 rows and 4 columns. As a result, each of the two images TI and CI has a total of 12 blocks B1 to B12. It is divided into blocks (step S1302). In the present embodiment, the main control unit 110 represents RGB values representing the representative colors of the blocks B1 to B12 of the two images TI and CI by combining red (R), green (G), and blue (B). Are extracted as feature amounts Ft and Fc (step S1304). The extracted feature amounts Ft and Fc are expressed using the following Equation 3.

  Here, the RGB value (Rti, Gti, Bti) represents the representative color of the “i” -th block Bi of the image TI, and the RGB value (Rci, Gci, Bci) represents the “i” -th block of the image CI. The representative color of Bi is shown. That is, the value of “n” in Equation 3 is the total number of divided blocks, and “n = 12” in the example shown in FIG. In this embodiment, Rti indicating a red value, Gti indicating a green value, and Bti indicating a blue value each take a value from “0” to “255”.

  Returning to the description of FIG. 4, after the feature amounts Ft and Fc are extracted (step S130), the main control unit 110 extracts the feature amount Fc extracted from the darker image CI of the two images TI and CI. In order to perform correction based on the evaluation values Lt and Lc extracted from the two images TI and CI, a feature amount correction process is executed (step S140).

  FIG. 9 is a flowchart showing the feature amount correction process (step S140). When starting the feature amount correction processing of FIG. 9, the main control unit 110 determines whether or not the value of ΔL, which is the difference between the evaluation value Lt and the evaluation value Lc, is smaller than the reference value Tk (step S1401). ΔL is expressed by the following Equation 4. In this embodiment, since the evaluation value Lc of the darker image CI is larger than the evaluation value Lt of the brighter image TI, the value of “Lt−Lc” is negative.

  When the value of ΔL is smaller than the reference value Tk (step S1401), the main control unit 110 ends the feature amount correction process of FIG. 9 without correcting the feature amount Fc.

  On the other hand, when the value of ΔL is equal to or greater than the reference value Tk (step S1401), the main control unit 110 calculates the correction value K based on the evaluation values Lt and Lc (step S1402). The correction value K is expressed by the following formula 5.

  After the correction value K is calculated (step S1402), the main control unit 110 gamma-corrects each RGB value (Rti, Gti, Bti) of the feature amount Fc using the following formula 6 (step S1404). The gamma value γ used for gamma correction may be a value included in the image file 710 of the image CI, or may be a preset value.

  After the feature value Fc is gamma-corrected (step S1404), the main control unit 110 converts the values (Rci ′, Gci ′, Bci ′) calculated by Expression 6 into correction values K using Expression 7 below. Based on this, level correction is performed (step S1405).

  After the level of the feature amount Fc is corrected (step S1405), the main control unit 110 reverses the values (Rci ″, Gci ″, Bci ″) calculated by Expression 7 using the following Expression 8. Gamma correction is performed (step S1406). Note that the gamma value γ in Expression 8 is the same value as the gamma value γ in Expression 5.

  After the feature value Fc is subjected to inverse gamma correction (step S1406), the main control unit 110 calculates whether or not the value of the feature value Fc corrected with the correction value K exceeds the maximum value, that is, Expression 8. It is determined whether at least one of the obtained values Rci ′ ″, Gci ′ ″, and Bci ′ ″ exceeds the maximum value “255” (step S1408). When the corrected feature value Fc exceeds the maximum value (step S1408), the main control unit 110 corrects (clipping) the value exceeding the maximum value “255” to “255”. The feature amount correction process 9 is ended. When the corrected feature value Fc does not exceed the maximum value (step S1408), the main control unit 110 ends the feature value correction process of FIG. 9 as it is.

  Returning to the description of FIG. 4, after the feature amount Fc is level-corrected (step S <b> 140), the main control unit 110 executes similarity determination processing for classifying the two images TI and CI into groups of similar images. (Step S150).

  FIG. 10 is a flowchart showing the similarity determination process (step S150). When starting the similarity determination process of FIG. 10, the main control unit 110 reads the flash information 7119 from each of the image files 710 in which the two images PI are recorded (step S1505), and based on each read flash information 7119, It is determined whether one of the two images TI and CI is photographed with a flash and the other is photographed without a flash (step S1510).

  If it is determined that the two images TI and CI are not in the presence / absence of flash (step S1510), the main control unit 110 sets the feature amount distance D, which is an index for determining the degree of similarity between the two images TI and CI. It calculates using the following numerical formula 9 (step S1520).

“Di” in Expression 9 represents a distance between the feature value Ft and the feature value Fc in the RGB Euclidean space. In this embodiment, the value of “s” in Equation 9 is “3 × 32 ² ”.

  After the feature amount distance D is calculated (step S1520), the main control unit 110 determines that the two images TI and CI are similar when the feature amount distance D is equal to or less than the reference value Td (step S1522). On the other hand, when the feature amount distance D is larger than the reference value Td (step S1522), it is determined that the two images TI and CI are dissimilar (step S1526).

  On the other hand, when it is determined that the two images TI and CI have a flash presence / absence relationship (step S1510), the main control unit 110 determines a feature amount distance D that is an index for determining the degree of similarity between the two images TI and CI. 'Is calculated using the following Equation 10 (step S1530).

“Di” in Expression 10 represents an approximate value Fapx that approximates the feature value Fc within the range of an assumed value Fsi that is assumed to change the feature value Ft when the image TI is captured without a flash, and a feature value Fc. Represents the distance between. Details of the feature distance D ′ will be described later. In this embodiment, the value of “s” in Equation 10 is “3 × 32 ² ”, which is the same as the value of “s” in Equation 9.

  After the feature amount distance D ′ is calculated (step S1530), the main control unit 110 determines that the two images TI and CI are similar when the feature amount distance D ′ is equal to or smaller than the reference value Td ′ (step S1532). On the other hand, if the feature distance D ′ is larger than the reference value Td ′ (step S1532), it is determined that the two images TI and CI are dissimilar (step S1536). . In the present embodiment, different threshold values are set for the reference value Td for determining the feature amount distance D and the reference value Td 'for determining the feature amount distance D'.

  Here, the details of the feature amount distance D ′ in Expression 10 will be described. FIG. 11 is an explanatory diagram showing a range of an assumed value Fsi that is assumed to change the feature amount Ft in the RGB Euclidean space. FIG. 11 shows a three-dimensional RGB Euclidean space including an R axis indicating the red component value, a G axis indicating the green component value, and a B axis indicating the blue component value. In the present embodiment, the assumed value Fsi is assumed that the subject itself does not change in color depending on the presence or absence of the flash, and the subject is brightly photographed by the flash. That is, the relationship between the feature amount Fti (Rti, Gti, Bti) and the assumed value Fsi (Rsi, Gsi, Bsi) is assumed to have the relationship of the following formulas 11 and 12.

  Therefore, as shown in FIG. 11, the assumed value Fsi is a value located on a straight line Lti that connects the feature amount Fti and the origin in the RGB Euclidean space.

  FIG. 12 is an explanatory diagram showing the distance | di | in the case where “Rti ≧ Rci and Gti ≧ Gci and Bti ≧ Bci” in Expression 10 is satisfied. In a case where “Rti ≧ Rci and Gti ≧ Gci and Bti ≧ Bci” is satisfied, the distance | di | Is the distance between. That is, the approximate value Fapx in this case is a point where a perpendicular line drawn from the feature amount Fci to the straight line Lti intersects the straight line Fci in the RGB Euclidean space.

  FIG. 13 is an explanatory diagram showing a distance | di | in a case where “Rti ≧ Rci and Gti ≧ Gci and Bti ≧ Bci” in Expression 10 is not satisfied. When “Rti ≧ Rci and Gti ≧ Gci and Bti ≧ Bci” are not satisfied, the distance | di | is the distance between the feature amount Fti and the feature amount Fci in the RGB Euclidean space, as in the case of Expression 9. . That is, the approximate value Fapx in this case is the same value as the feature amount Fti.

  Although it is desirable that the feature amount distance D ′ is calculated based on the approximate value Fapx obtained on the basis of the feature amount Ft with flash as in the present embodiment, it is obtained on the basis of the feature amount Fc without flash. The feature amount distance D ′ may be calculated based on the obtained approximate value Fapx. When the feature amount Ft with flash is used as a reference, the error of the approximate value Fapx can be made smaller than when the feature amount Fc without flash is used as a reference. FIG. 14 is an explanatory diagram showing an error of the approximate value Fapx. FIG. 14 shows the error Eft of the feature amount Ft with flash and the error Efc of the feature amount Fc without flash on a two-dimensional Euclidean space composed of the R axis and the G axis. As shown in FIG. 14, even if the error Eft with flash and the error Efc without flash are the same size, the approximate value error Eat, which is the error of the approximate value Fapx with the flash reference, is the approximate value with the non-flash reference. It becomes smaller than the approximate value error Eac which is an error of Fapx. In FIG. 14, a two-dimensional Euclidean space composed of an R axis and a G axis is shown for simplicity of explanation, but the same applies to a three-dimensional Euclidean space.

  Returning to the description of FIG. 4, after the similarity between the two images TI and CI is determined (step S150), the main control unit 110, when the two images TI and CI are similar (step S152), The two images TI and CI are classified into the same group (step S154). On the other hand, if the two images TI and CI are dissimilar (step S152), the two images TI and CI are put into different groups, respectively. Classification is performed (step S156).

  FIG. 15 is an explanatory diagram showing how the plurality of images PI in FIG. 4 are grouped (steps S152, S154, S156). In this embodiment, the similarity is sequentially determined for the images PI that are adjacent to each other in the order in which the images were taken. Therefore, the images for each group are consecutive in the order in which the images were taken. It becomes a boundary between them. In the example shown in FIG. 11, among the plurality of images PI arranged in the shooting order from the left, the first to third images PI from the left are determined to be similar, and the third and fourth are determined to be dissimilar. The state in which the th and fifth images PI are determined to be similar is shown. As a result, in the example shown in FIG. 11, the first to third images PI are classified into the group G1, and the fourth and fifth images PI are classified into the group G2.

  Returning to the description of FIG. 4, after grouping all the images (step S158), the main control unit 110 selects at least one image PI from each group into which the plurality of images PI are classified (step S160). Then, control is performed to print the selected image PI (step S170), and the image printing process in FIG. 4 is terminated. In this embodiment, in selecting an image PI from each group (step S160), the main control unit 110 extracts each edge amount of the image PI belonging to each group, and selects an image PI having a large edge amount in each group. Select one. In the present embodiment, if there is one image PI belonging to the group, that one image PI is selected.

  According to the printing apparatus 10 described above, since the influence of the presence or absence of flash on the determination of similarity between images can be reduced (similarity determination processing in FIG. 10), determination of similarity between images is appropriate. Can be done. As a result, it is possible to appropriately execute image printing according to the similarity. Further, since the feature amount distance D ′ is calculated on the basis of the feature amount Ft with flash (Formula 10 and FIG. 14), the similarity between images can be determined more appropriately.

B. Other embodiments:
As mentioned above, although the Example of this invention was described, this invention is not limited to the above-mentioned embodiment at all, Of course, it can implement with various forms within the range which does not deviate from the meaning of this invention. It is. For example, in the above-described embodiment, the correction value K for correcting the feature amount F is a value based on the photographing condition of the image PI, but may be a value based on the luminance of the image PI itself. FIG. 16 is a flowchart showing the evaluation value extraction process (step S125) in the modification. The main control unit 110 in the modified example executes the evaluation value extraction process (step S125) in FIG. 16 instead of the evaluation value extraction process (step S120) in FIG. 6 in the image printing process in FIG.

  When starting the evaluation value extraction process of FIG. 16, the main control unit 110 creates luminance histograms LHt and LHc from each of the two images TI and CI (step S1252). FIG. 17 is an explanatory diagram showing a state of the evaluation value extraction process (step S125) of FIG. FIG. 17 shows luminance histograms LHt and LHc created from each of the two images TI and CI. The luminance histograms LHt and LHc in FIG. 17 are graphs in which the luminance values of the pixels constituting each image are taken on the horizontal axis and the cumulative frequency of the pixels constituting the image is taken on the vertical axis. After the luminance histograms LHt and LHc are created (step S1252), the main control unit 110 calculates the luminance value taking the cumulative frequency NA in the luminance histograms LHt and LHc as the evaluation value Y (step S1254). In this modification, the cumulative frequency NA used for calculating the evaluation value Y is “10%”. In this modification, of the two extracted evaluation values Y, the larger one is expressed as an evaluation value Yt, the smaller one is expressed as an evaluation value Yc, and the brighter one, that is, the evaluation value Yt. Is extracted as an image TI, and a darker one, that is, an evaluation value Yc is extracted as an image CI.

  In the present modification, in the feature amount correction process (step S140) in FIG. 9, the main control unit 110 calculates the correction value K using the following equation 13 (step S1402).

  According to the printing apparatus 10 of the modified example described above, the influence of the difference in brightness of the entire image PI between the images PI on the determination of the similarity between the images PI is reduced based on the luminance of each image PI itself. be able to.

  9, the main control unit 110 omits the gamma correction (step S1404) and the inverse gamma correction (step S1406), and corrects the feature quantity Fc using the correction value K. You may carry out. As a result, it is possible to increase the processing speed in the image PI analogy determination.

  Further, the main control unit 110 uses gamma correction (step S1404), level correction (step S1405), and inverse gamma in the feature amount correction processing of FIG. Each process of correction (step S1406) may be performed collectively. As a result, it is possible to increase the processing speed in the image PI analogy determination.

  In the above-described embodiment, the feature quantity Fc of the two feature quantities Ft and Fc is corrected based on the evaluation values Lt and Lc. However, the feature quantity Ft is corrected instead of the feature quantity Fc. Alternatively, both of the two feature amounts Ft and Fc may be corrected.

  In the above-described embodiment, the determination of the similarity between images is performed for grouping a plurality of images PI, but may be performed for searching for similar images from the plurality of images PI. FIG. 18 is a flowchart showing a similar image search process in the modification. In the main control unit 110 of the printing apparatus 10, the storage medium 810 and the external device 820 are attached to the card slot 140 and the communication connector 150 by the user, and an execution request for the image search function is input by the user via the input button unit 170. At this time, the similar image search process of FIG. 18 is started. The similar image search process of FIG. 18 determines the similarity between the image PI selected by the user from the plurality of images PI and the other images PI (steps S210, S215, S258), and the similar images searched Except for displaying the PI as a search result (step S270), the process is the same as the image printing process of FIG. Further, at least one of the images PI retrieved by the similar image retrieval process of FIG. 18 may be selected and printed.

  In the above-described embodiment, the printing apparatus 10 to which the present invention is applied has been described. However, other apparatuses that determine the similarity between images in a plurality of images, such as a personal computer, a digital still camera, a photo viewer, and the like. In addition, the present invention may be applied.

1 is a perspective view illustrating an external configuration of a printing apparatus 10. 2 is an explanatory diagram mainly showing an internal configuration of a printing apparatus 10; FIG. It is explanatory drawing which shows the data structure of the image file 710 typically. 4 is a flowchart illustrating image printing processing executed by the main control unit 110 of the printing apparatus 10. FIG. 5 is an explanatory diagram showing how two images PI are specified in the image printing process of FIG. 4 (step S110). It is a flowchart which shows an evaluation value extraction process (step S120). It is a flowchart which shows a feature-value extraction process (step S130). It is explanatory drawing which shows the mode of the feature-value extraction process (step S130) of FIG. It is a flowchart which shows a feature-value correction process (step S140). It is a flowchart which shows the similarity determination process (step S150). It is explanatory drawing which shows the range of the assumption value Fsi assumed that the feature-value Ft changes in RGB Euclidean space. It is explanatory drawing which shows distance | di | in the case of satisfy | filling "Rti> = Rci and Gti> = Gci and Bti> = Bci" in Numerical formula 10. It is explanatory drawing which shows distance | di | when not satisfy | filling "Rti> = Rci and Gti> = Gci and Bti> = Bci" in Numerical formula 10. It is explanatory drawing which shows the error of approximate value Fapx. FIG. 5 is an explanatory diagram showing a state (steps S152, S154, S156) of a plurality of images PI in FIG. It is a flowchart which shows the evaluation value extraction process (step S125) in a modification. It is explanatory drawing which shows the mode of the evaluation value extraction process (step S125) of FIG. It is a flowchart which shows the similar image search process in a modification.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Printing apparatus 110 ... Main control part 111 ... Image reading part 112 ... Image specific part 113 ... Feature-value extraction part 114 ... Evaluation value extraction part 115 ... Feature-value correction part 116 ... Similarity determination part 117 ... Image selection part 118 ... Print control unit 120 ... printing unit 130 ... scanner unit 140 ... card slot 150 ... communication connector 160 ... display unit 170 ... input button unit 200 ... carriage 210 ... print head unit 211-216 ... ink ejection head 240 ... carriage drive unit 250 ... Print medium transport unit 710 ... Image file 7110 ... Shooting condition data 7112 ... Shooting date and time 7114 ... Aperture value 7116 ... Shutter speed 7118 ... ISO sensitivity 7119 ... Flash information 7120 ... Image data 810 ... Storage medium 820 ... External device 900 ... Print medium G1 , 2 ... Group PI, TI, CI ... Image Bi ... Block F, Ft, Fc ... Feature quantity D, D '... Feature quantity distance Fsi ... Assumed value Fapx ... Approximate value L, Lt, Lc ... Evaluation value [Delta] L ... Difference K ... Correction value Y, Yt, Yc ... Evaluation value NA ... Cumulative frequency

Claims (15)

A printing apparatus for printing some of a plurality of images stored in a storage medium,
An image reading unit for reading the plurality of images from the storage medium;
An image specifying unit for specifying two images from the plurality of read-out images;
A feature amount extracting unit that extracts a feature amount indicating a color feature constituting the image from each of the two specified images;
A similarity determination unit that determines the similarity between the two images based on the two extracted feature quantities;
An image selection unit that selects an image from the plurality of images based on the determined similarity;
A print control unit that performs control to print the selected image,
The similarity determination unit
A flash determination unit that determines whether one of the two identified images is captured with a flash and the other is captured without a flash;
When it is determined that there is a flash presence / absence relationship, an approximate value that approximates the other of the feature amounts within a range of values that one of the two feature amounts is assumed to change depending on the presence / absence of flash, A printing apparatus comprising: a feature amount replacement unit that determines the similarity by treating the feature amount as one of the feature amounts.   The printing apparatus according to claim 1, wherein one of the feature amounts is a feature amount extracted from an image photographed with a flash.   The printing apparatus according to claim 1, wherein the assumed value is a value that indicates a characteristic of a color that is the same color as that indicated by one of the feature amounts and has a different brightness.   The storage medium stores, for each image, flash information indicating the presence / absence of a flash when the image is captured, and the flash determination unit has a relationship of presence / absence of the flash based on the flash information. The printing apparatus according to claim 1, wherein it is determined whether or not.   5. The printing apparatus according to claim 4, wherein the flash information is Exif standard information. The printing apparatus according to claim 1, further comprising:
An evaluation value extracting unit that extracts an evaluation value indicating the brightness of the image from each of the two specified images;
A printing apparatus comprising: a feature amount correction unit that corrects at least one of the extracted two feature amounts in a direction in which the brightness of the two images approaches based on the extracted evaluation value. The printing apparatus according to any one of claims 1 to 6,
The feature amount extraction unit includes:
A block unit that divides each of the two identified images into a plurality of blocks;
An RGB extraction unit that extracts an RGB value indicating a representative color for each of the plurality of blocks as the feature amount;
The similarity determination unit determines the similarity between the two images based on the distance in the RGB Euclidean space for the two RGB values respectively extracted from the corresponding blocks between the two images. . The printing apparatus according to any one of claims 1 to 7,
The image selection unit
An image classifying unit that classifies the plurality of images into groups for each similar image based on the similarity between the determined two images;
A printing apparatus comprising: an image selection unit that selects, as the selected image, at least one image included in the group for each classified group.   The plurality of images stored in the storage medium are images arranged in a series in the order of shooting, and the two specified images are images that are adjacent in the order of shooting. Printing device. The printing apparatus according to any one of claims 1 to 7,
The image selection unit
An image search unit that searches the plurality of images for an image similar to one of the specified images based on the determined similarity between the two images;
A printing apparatus comprising: an image selection unit that selects at least one of the searched images as the selected image. An image processing apparatus that processes a plurality of images stored in a storage medium,
An image reading unit for reading the plurality of images from the storage medium;
An image specifying unit for specifying two images from the plurality of read-out images;
A feature amount extracting unit that extracts a feature amount indicating a color feature constituting the image from each of the two specified images;
A similarity determination unit that determines the similarity between the two images based on the two extracted feature quantities;
The similarity determination unit
A flash determination unit that determines whether one of the two identified images is captured with a flash and the other is captured without a flash;
When it is determined that there is a flash presence / absence relationship, an approximate value that approximates the other of the feature amounts within a range of values that one of the two feature amounts is assumed to change depending on the presence / absence of flash, An image processing apparatus comprising: a feature amount replacement unit that determines the similarity by handling the feature amount as one of the feature amounts.   The image processing apparatus according to claim 11, further comprising an image classification unit that classifies the plurality of images into groups for each similar image based on the determined similarity between the two images.   The image processing according to claim 11, further comprising: an image search unit that searches the plurality of images for an image similar to one of the identified images based on the determined similarity between the two images. apparatus. An image processing method for handling a plurality of images stored in a storage medium,
Reading the plurality of images from the storage medium;
Two images are identified from the plurality of read-out images,
Extracting a feature amount indicating a color feature constituting the image from each of the two specified images;
Determining the similarity of the two images based on the two extracted feature quantities;
When determining the similarity between the two images, it is determined whether one of the two specified images is taken with a flash and the other is taken without a flash.
When it is determined that there is a flash presence / absence relationship, an approximate value that approximates the other of the feature amounts within a range of values that one of the two feature amounts is assumed to change depending on the presence / absence of flash, An image processing method for determining the similarity by treating it as one of feature quantities. A program for causing a computer to execute a function of processing a plurality of images,
A function of reading the plurality of images from the storage medium;
A function of specifying two images from the plurality of read-out images;
A function for extracting a feature amount indicating a color feature constituting the image from each of the two specified images;
A function of determining the similarity between the two images based on the two extracted feature values;
A function for determining whether or not there is a relationship between the presence or absence of a flash in which one of the two specified images is taken with a flash and the other is taken without a flash when judging the similarity between the two images When,
When it is determined that there is a flash presence / absence relationship, an approximate value that approximates the other of the feature amounts within a range of values that one of the two feature amounts is assumed to change depending on the presence / absence of flash, A program for realizing the function of determining the similarity by treating it as one of the feature quantities.

Images