JP2007011939A - Image decision device and method therefor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image decision device and a method therefor, capable of automatically deciding whether a scene of an image is an artifact image obtained by photographing a building or the like or not, not through an operator. <P>SOLUTION: This image decision device deciding the scene of the image has: an image conversion means 12 converting the image that is a decision target into a gray scale image; a gradient intensity calculation means 13 calculating a gradient intensity and a gradient direction of each pixel of the converted gray scale image; an edge image generation means 14 generating an edge image wherein the gray scale image is binarized on the basis of the calculated gradient intensity; a characteristic amount derivation means 15 extracting a line segment configured by continuous pixels in a prescribed range included in the generated edge image, and deriving a characteristic amount thereof; and a scene decision means 16 deciding the scene of the image from the derived characteristic amount. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image determination apparatus and method for determining an image scene.

Conventionally, in a photographic image processing apparatus, correction processing according to the scene of a photographic image has been performed. For example, in the case of an artifact image in which a building is photographed, a sharp image with enhanced edges and contrast is generally preferred, so that correction processing such as sharpening processing is performed, In the case of a natural image in which a natural object is photographed, generally, sharpening processing or the like is not performed.
JP 2000-270222 A

  However, until now, some scenes in photographic images have been proposed with algorithms such as human detection, but it is determined whether or not a building or the like is an image of an artificial object. In this case, it is necessary to make a determination for each photographic image by the operator's confirmation work, and it has been a problem that a lot of time is spent on the correction processing according to the scene of the photographic image.

  In view of the above-described conventional problems, the present invention provides an image determination device capable of automatically determining whether an image scene is an artifact image obtained by photographing a building or the like without using an operator, and The point is to provide such a method.

  In order to achieve the above object, a first characteristic configuration of an image determination device according to the present invention is an image determination device for determining a scene of an image as described in claim 1 of the document of the claim, Based on the calculated gradient strength, image conversion means for converting the image to be determined into a grayscale image, gradient strength calculation means for calculating the gradient strength and gradient direction of each pixel of the converted grayscale image, and Edge image generation means for generating an edge image obtained by binarizing an image, and feature quantity derivation means for extracting a line segment composed of a predetermined range of continuous pixels included in the generated edge image and deriving the feature quantity And a scene determination means for determining the scene of the image from the derived feature amount.

  In various scenes in an image, for example, there are scenes including many long and straight line segments as represented by an artifact image in which a building or the like is photographed. In such a case, according to the above-described configuration, the scene determination unit uses the feature amount of the line segment including the predetermined range of continuous pixels included in the edge image based on the gradient strength in the grayscale image to generate an image scene. Therefore, it is possible to determine whether or not the scene includes many long straight line segments.

  In the second feature configuration, as described in claim 2, in addition to the first feature configuration described above, the gradient strength calculation means includes a gradient strength, a vertical direction, a horizontal direction in a pixel of the grayscale image. The point is that the gradient direction is calculated in the four directions of the direction and the 45-degree tilt direction.

  With the above configuration, a line segment composed of a predetermined range of continuous pixels included in the edge image is specified as a line segment in four directions of the vertical direction, the horizontal direction, and the 45-degree tilt direction of the pixel of the edge image. Therefore, it is possible to efficiently determine whether or not the scene includes a geometric line segment peculiar to an artifact among scenes including many long straight line segments. It becomes.

  In the third feature configuration, as described in claim 3, in addition to the first or second feature configuration described above, the edge image generation means converts the line segment included in the generated edge image into the gradient. There is a thinning processing means for selecting based on the direction and the gradient intensity.

  According to the above configuration, in order to select the line segment included in the edge image generated by the thinning processing unit based on the gradient direction and the gradient intensity, for example, linear lines adjacent to several rows in the same direction in the edge image. A thin line segment can be thinned to form a line segment. In addition, an edge image based on the gradient strength of a grayscale image in an image that is slightly thick and gently curved may be an edge image in which two line segments having different lengths are adjacent to each other. According to the above-described configuration, in such a case, each of the two line segments can be divided into short lines and can be made into one line for each of the divided line segments. That is, it is possible to prevent an edge image based on the gradient strength of the grayscale image in an image that is slightly thick and gently curved, which is unique to natural objects, from becoming a long and straight line segment.

  In the fourth feature configuration, in addition to any one of the first to third feature configurations described above, the feature quantity deriving unit may calculate the average length of the extracted line segment. The standard deviation is derived as a feature quantity.

  With the above-described configuration, the scene determination means can determine the average length of the line segment as an indicator of whether or not it is a long straight line segment unique to the artifact, It is possible to determine the scene of the image based on the standard deviation as an index of whether or not it is a geometric line segment. In other words, the scene determination means can determine whether or not the scene includes many long straight line segments, and whether or not the scene includes a geometric line segment. It is.

  In the fifth feature configuration, as described in claim 5, in addition to the fourth feature configuration described above, the feature amount deriving means may include a predetermined maximum number of pixels constituting a line segment in each direction. A predetermined number of line segments are extracted in the order of the number of pixels that is smaller than the number of pixels, and the average of the average values in the respective directions of the extracted line segments is derived as the average length.

  According to the above-described configuration, the feature amount deriving unit extracts a line segment having a line segment that is smaller than a predetermined maximum number of pixels in each direction. For example, a line based on a natural object such as a horizon line is used. The feature amount deriving means extracts a predetermined number of line segments in descending order of the number of pixels constituting the line segment in each direction, and calculates an average value of the extracted line segments in each direction. Since the average is derived as the average length, the average length can be set as the average length as the feature amount of the line segment based on the artifact.

  In the sixth feature configuration, as described in claim 6, in addition to any one of the first to fifth feature configurations described above, the scene determination unit may include a Mahalanobis general distance using the feature amount as a variable. This is because the scene of the image is determined by discriminant analysis based on the above, and the scene of the image can be reliably determined based on a preset sample.

  In the seventh feature configuration, as described in claim 7, in addition to any of the first to sixth feature configurations described above, the scene determination means captures an image to be determined as a natural object. Whether it is a natural image or a man-made image of the building, and whether it is a scene with many long straight lines, and whether it is a geometric It can be determined whether or not the scene is composed of straight line segments as whether or not the building is a man-made object image.

  In order to achieve the above object, a first characteristic configuration of an image determination method according to the present invention is an image determination method for determining a scene of an image as described in claim 8, wherein an image to be determined is selected. An image conversion step for converting to a grayscale image, a gradient strength calculation step for calculating the gradient strength and gradient direction of each pixel of the converted grayscale image, and binarization of the image based on the calculated gradient strength An edge image generating step for generating an edge image, a feature amount deriving step for extracting a line segment composed of continuous pixels within a predetermined range included in the generated edge image, and deriving the feature amount; and a derived feature And a scene determination step for determining an image scene from the quantity.

  According to the second feature configuration, in addition to the first feature configuration described above, in the gradient strength calculation step, the gradient strength at the pixel of the grayscale image, the vertical direction, the horizontal direction, and the like. It is in the point which calculates the direction and the gradient direction of 4 directions of a 45-degree inclination direction.

  In the third feature configuration, in addition to the first or second feature configuration described above, in the edge image generation step, a line segment included in the generated edge image is converted into the gradient. The thinning process is selected based on the direction and the gradient strength.

  In the fourth feature configuration, as described in claim 11, in addition to any one of the first to third feature configurations described above, in the feature amount derivation step, The standard deviation is derived as a feature quantity.

  In the fifth feature configuration, as described in claim 12, in addition to the fourth feature configuration described above, in the feature amount deriving step, the number of pixels constituting a line segment in each direction is a predetermined maximum pixel. A predetermined number of line segments are extracted in the order of the number of pixels that is smaller than the number of pixels, and the average of the average values in the respective directions of the extracted line segments is derived as the average length.

  In the sixth feature configuration, as described in claim 13, in addition to any of the first to fifth feature configurations described above, in the scene determination step, the Mahalanobis general distance using the feature amount as a variable. The scene of the image is determined by discriminant analysis based on the above.

  In the seventh feature configuration, as described in claim 14, in addition to any of the first to sixth feature configurations described above, in the scene determination step, a natural object is photographed as an image to be determined. In other words, it is determined whether the image is a natural image or an artifact image obtained by photographing a building.

  As described above, according to the present invention, it is possible to automatically determine whether an image scene is an artifact image obtained by photographing a building or the like, without using an operator, and a method thereof. Can now be provided.

  An embodiment of a photographic image processing apparatus incorporating an image determination apparatus according to the present invention will be described below. As shown in FIG. 2, the photographic image processing apparatus 1 includes a photographic printer 2 that performs an exposure process on the photographic paper P based on output image data and develops the exposed photographic paper, and a developed photographic film F. A media driver 32 that reads image data from an image data storage medium M such as a memory card that stores image data taken by a film scanner 31 or a digital still camera that reads an image from the camera, a general-purpose computer as a controller 33, and the like The print data is set and inputted with the print order information for the inputted photographic image as the original image, and includes an operation station 3 for performing various image correction processes, and the print data edited from the original image by the operation station 3 Is output to the photographic printer 2 to produce a desired photographic print. That.

  As shown in FIGS. 2 and 3, the photographic printer 2 has two systems of photographic paper magazines 21 containing roll-shaped photographic paper P, and photographic paper P drawn from the photographic paper magazine 21 with a predetermined print. A sheet cutter 22 that cuts into a size; a back print unit 23 that prints print information such as a frame number on the back of the cut photographic paper P; an exposure unit 24 that exposes the photographic paper P based on the print data; A development processing unit 25 including a plurality of processing tanks 25a, 25b, and 25c filled with processing solutions for developing, bleaching, and fixing the exposed photographic paper P is disposed along the transport path of the photographic paper P. The laterally-feeding conveyor 26 that discharges the photographic paper P that has been dried after the development process, and the sheet-paper (photo print) P that is stacked on the laterally-feeding conveyor 26 is sorted in order units. Configured to include the data 27.

  The exposure unit 24 receives a laser beam bundle of RGB three colors modulated based on the print data in the main scanning direction orthogonal to the conveyance direction with respect to the photographic paper P conveyed in the sub-scanning direction by the conveyance mechanism 28. An exposure head 24a for outputting and exposing is accommodated.

  A transport mechanism 28 including a plurality of roller pairs that transport the photographic printing paper P at a process speed corresponding to the exposure unit 24 and the development processing unit 25 disposed along the transport path is disposed before and after the exposure unit 24. Is provided with a chucker-type transport mechanism 28a capable of transporting photographic paper P in multiple rows.

  The controller 33 provided in the operation station 3 is installed with an application program that operates under the control of a general-purpose operating system and executes various controls of the photographic processing apparatus 1, and a monitor 34 as an operation interface with the operator. A keyboard 35, a mouse 36, and the like are connected.

  The photographic processing process executed by the cooperation of the hardware and software of the controller 33 will be described in functional blocks. As shown in FIG. 4, photographic image data read by the film scanner 31 and the media driver 32 is received. An image input unit 40 that performs predetermined preprocessing and transfers it to a memory 41, which will be described later, and print order information and image editing information are displayed on the screen of the monitor 34, and an operation for inputting necessary data for them. A graphic user interface unit (GUI) 42 for generating a graphic operation screen for displaying a user icon or generating various control commands based on input operations from the keyboard 35 and mouse 36 to the displayed graphic operation screen; Transfer from the image input unit 40 And a memory 41 in which photographic image data to be processed, photographic image data after correction processing by an image processing unit 47 to be described later, correction parameters at that time, and set print order information, etc. are partitioned and stored in a predetermined area; An order processing unit 43 that generates order information, an image processing unit 47 that performs density correction processing and contrast correction processing on each photographic image data stored in the memory 41, and a display from the graphic user interface unit 42 A display control unit 46 including a video RAM for displaying the image data and various input / output graphic data developed in the memory 41 based on the command on the monitor 34, and the final of the various correction processes. A pre-image for generating print data for outputting the corrected image to the photo printer 2 And Todeta generator 44, and the like formatter 45 which converts the final corrected image according to customer orders to a file format for writing in a storage medium such as a CD-R.

  The film scanner 31 is configured to operate in two modes: a pre-scan mode that reads an image recorded on the film F at a high speed although it has a low resolution, and a main scan mode that reads a high-resolution image at a low speed. Various correction processes are performed in the pre-judge mode on the low-resolution image read in the can mode, and the high-resolution image read in the main scan mode based on the correction parameters stored in the memory 41 at that time. The final correction process for the image is executed and output to the printer 2.

  Similarly, the image file read from the media driver 32 includes a high-resolution captured image and its thumbnail image, and various correction processes described later are performed on the thumbnail image. Based on the stored correction parameters, a final correction process is performed on the high-resolution captured image. When the image file does not include a thumbnail image, the image input unit 40 generates a thumbnail image from the high-resolution captured image and transfers it to the memory 41. In this way, the calculation load of the controller 33 is reduced by executing various editing processes that are frequently trial and error on low-resolution images.

  As shown in FIG. 1, the image processing unit 47 includes an input image determination unit 10 that is a main part of the present invention that determines a scene of an input image, and an image whose scene is determined by the input image determination unit 10. On the other hand, based on the determination result or based on an operation instruction by the operator from the keyboard 35 or mouse 36 via the GUI 42, image density correction processing, image contrast correction processing, and image edge enhancement are emphasized. And correction processing means 11 for performing correction processing such as sharpening processing for reducing noise.

  The input image determination means 10 calculates an image conversion means 12 for converting an image to be determined into a grayscale image, and a gradient strength and a gradient direction of each pixel of the grayscale image converted by the image conversion means 12. Gradient intensity calculating means 13, edge image generating means 14 for generating an edge image obtained by binarizing the image based on the gradient intensity calculated by the gradient intensity calculating means 13, and the edge image generating means 14 A feature amount deriving unit 15 for extracting a line segment composed of continuous pixels within a predetermined range included in the edge image and deriving the feature amount; and a scene of the image from the feature amount derived by the feature amount deriving unit 15. A scene determination means 16 for determination is provided.

  For example, the image conversion unit 12 converts each pixel data of the image to be determined into a value obtained by weighting and adding each of the R component, the G component, and the B component based on [Equation 1]. Thus, the image to be determined is converted to a gray scale image.

  Y (i) is rounded to an integer, and i indicates each pixel number.

  The gradient intensity calculating means 13 calculates the gradient intensity and gradient direction of each pixel of the grayscale image converted by the image converting means 12, and has, for example, a gradient intensity as shown in FIG. A vertical component P1 (i) of the gradient intensity at each pixel of the grayscale image is calculated by multiplying each pixel data of the grayscale image by a Sobel filter that extracts a vertical component, and FIG. By multiplying each pixel data of the gray scale image by a Sobel filter for extracting the horizontal component of the gradient intensity as shown in (b), the horizontal component P2 of the gradient intensity at each pixel of the gray scale image ( i) is calculated, and further, gradient intensity Pa (i) at each pixel of the gray scale image is calculated based on [Equation 2], and based on [Equation 3]. Is configured to calculate the gradient direction Sa (i) in each pixel of the grayscale image are.

  The gradient direction Sa (i) is configured to be classified into four directions, that is, a vertical direction and a horizontal direction of pixels of the gray scale image, and a 45-degree inclined direction intersecting each other. For example, when −90.0 ≦ Sa (i) <− 67.5 and 67.5 <Sa (i) ≦ 90.0, the vertical direction of the pixel of the gray scale image, −22.5 ≦ Sa (i) When ≦ 22.5, the horizontal direction of the pixels of the grayscale image, and when −67.5 ≦ Sa (i) <− 22.5, the 45 ° inclination direction of one of the pixels of the grayscale image is 22.5 ≦ When Sa (i) <67.5, the other 45 degree tilt direction of the pixels of the gray scale image can be classified.

  That is, the gradient intensity calculation means 13 sets the gradient intensity Pa (i) for each pixel of the grayscale image at 45 degrees that intersects the vertical direction, the horizontal direction, or the pixels of the grayscale image. The gradient strength is calculated as a gradient strength for any one of the four gradient directions.

  The edge image generation unit 14 detects, as edge pixels, pixels of the image in which the gradient intensity Pa (i) calculated by the gradient intensity calculation unit 13 is equal to or greater than a predetermined gradient intensity threshold Pth. A binarized edge image as shown in FIG. 6B is generated by dividing the detected edge pixel into other pixels from the image as shown in FIG. 6A. It is configured.

  The gradient intensity threshold value Pth is used as an edge pixel with respect to the number of pixels of the image, based on a gradient intensity histogram as shown in FIG. 7 obtained from the gradient intensity Pa (i) calculated by the gradient intensity calculating means 13. It can set suitably so that the ratio of the number of detected pixels may become a predetermined ratio. The gradient intensity threshold value Pth is preferably within a predetermined gradient intensity range. For example, when the grayscale image is 8-bit data, the gradient intensity threshold value Pth is at least in the range of 100 to 320. If it is set within the range, the influence of noise or the like can be reduced, which is preferable.

  Further, the edge image generation means 14 selects a line segment included in the generated binarized edge image based on the gradient direction and the gradient intensity by the thinning processing means 17 provided therein. It is configured.

  More specifically, the thinning processing unit 17 compares the gradient directions of the two edge pixels when the edge pixels are arranged in adjacent pixels in the binarized edge image. 8 (b), FIG. 8 (c), and FIG. 8 (d), the gradient direction is the same direction, and the two edge pixels are aligned in a direction parallel to the gradient direction. In such a case, by omitting the edge pixel having the smaller gradient strength from the two edge pixels from the edge pixel, the binarized edge image as shown in FIG. A thinned binarized edge image is generated as shown in FIG.

  That is, the thinning processing unit 17 is configured to select a line segment composed of edge pixels having a higher gradient strength when line segments as continuous edge pixels composed of edge pixels connected in the same direction are adjacent to each other. Has been.

  Here, as described above, the binarized edge image is generated based on the gradient strength. Therefore, since the gradient direction of each edge pixel in the line segment as the continuous edge pixel is mostly the same direction, as described above, the gradient direction is the same direction, and the two edge pixels are Only when they are arranged in a direction parallel to the gradient direction, the thinning process can be sufficiently performed by omitting predetermined edge pixels.

  Further, the thinning processing unit 17 generates, for example, an edge image in which two line segments having different lengths as shown in FIG. 10 generated based on an image that is slightly thick and gently curved are adjacent to each other. On the other hand, by performing the thinning process, each of the two line segments is shortened or divided, and further, the shortened or divided line segments are arranged in a line for each line segment, so that it becomes a natural object. The edge image based on the gradient strength of the grayscale image in the characteristic slightly thick and gently curved image is prevented from becoming a long and straight line segment.

  The feature quantity deriving unit 15 includes a binary edge image finally generated by the edge image generating unit 14, that is, a continuous edge pixel in a predetermined range included in the binarized edge image subjected to the thinning process. The line length is extracted and the feature amount is derived, and the run lengths are respectively obtained for the vertical direction, the horizontal direction, and the 45-degree inclined directions intersecting each other in the vertical direction and the horizontal direction of the pixels of the binarized edge image. By obtaining a matrix, a line segment length histogram (continuous edge pixel number histogram) showing the relationship between the length of each line segment and the number of line segments corresponding to that length as shown in FIG. 11 in the four directions. , And calculate the average line length from the line segment with the maximum line segment length to the line segment with the line segment length of the predetermined rank. The average value of The average value, is configured to derive the feature amount and the determination standard deviation is the standard deviation of the average segment length of the calculated.

  That is, the feature quantity deriving unit 15 determines whether the image includes a discrimination average value as an index as to whether or not the image includes a long straight line segment peculiar to the artifact, and the long straight line segment is geometric. A standard discriminant quasi-deviation as an index as to whether or not it is a straight line segment is derived as a feature quantity.

  The average line length is omitted from a line segment having a maximum line segment length below the predetermined line segment length threshold Cth by omitting a line segment length longer than the predetermined line segment length threshold Cth. If the average line length up to the line segment that is the rank line length is taken, for example, a line segment based on a natural object such as a horizon line can be omitted, and the average line segment length is more based on an artifact. The average line segment length as the characteristic amount of the line segment can be preferably used. For example, when the image has 256 × 384 pixels, the line segment length threshold Cth is preferably set to about 50, so that line segments based on natural objects such as the horizon can be omitted. When the average line segment length is the average value of the top 30 segment lengths, the image is applied as an index as to whether or not the image includes a long linear segment peculiar to artifacts. Suitable and preferred.

  The scene determination unit 16 determines an image scene from the discrimination average value and the discrimination standard deviation as the feature amount derived by the feature amount deriving unit 15, and preliminarily serves as sample values for various scenes. Several populations are set, for example, based on Mahalanobis's general distance, it is determined whether the derived discriminant average value and discriminant standard deviation belong to the population, and the discriminant average value and discriminant standard deviation The scene corresponding to the population determined to belong to is determined as the scene in the discrimination target image.

  Specifically, the scene determination means 16 includes a population consisting of a discrimination average value and a discrimination standard deviation as a plurality of sample values in an artifact image obtained by photographing one of the populations in advance for various buildings. Is set as a population composed of a discriminant average value and discriminant standard deviation as a plurality of sample values in images taken by other populations other than buildings, and the scene judgment means 16 is set to the Mahalanobis general distance. By determining the population to which the feature amount belongs, it is configured to determine whether or not the discrimination target image is an artifact image obtained by photographing a building.

  Hereinafter, the scene determination operation of the image will be described based on the flowchart of FIG. When an image is stored in the memory 41 (SA1), the image conversion means 12 weights each pixel data of the image for each of the R component, the G component, and the B component based on the [Equation 1]. The image is converted into a grayscale image by converting it into an added value of the obtained values (SA2).

  When the image stored in the memory 41 is converted into a grayscale image by the image converting means 12, the gradient strength calculating means 13 applies a Sobel filter for extracting the longitudinal component of the gradient strength to the grayscale image. By multiplying each pixel data, the vertical component P1 (i) of the gradient intensity at each pixel of the gray scale image is calculated (SA3). In addition, the horizontal component P2 (i) of the gradient intensity at each pixel of the grayscale image is calculated by multiplying each pixel data of the grayscale image by a Sobel filter that extracts the horizontal component of the gradient intensity ( SA4). Further, the gradient intensity Pa (i) at each pixel of the grayscale image is calculated based on [Equation 2] (SA5), and the gradient direction at each pixel of the grayscale image is calculated based on [Equation 3]. Sa (i) is calculated (SA6). At this time, the gradient direction Sa (i) is classified into four directions, ie, a vertical direction, a horizontal direction, and a 45-degree inclined direction of the pixels of the grayscale image.

  When the gradient strength calculation unit 13 calculates the gradient strength Pa (i) and the gradient direction Sa (i) in the grayscale image, the edge image generation unit 14 calculates the gradient strength Pa (( i), the gradient intensity histogram is generated (SA7), and a gradient intensity threshold value Pth at which a ratio of the number of pixels detected as edge pixels to the number of pixels of the image from the gradient intensity histogram is a predetermined ratio is set. Calculate (SA8). Then, a binarized edge image obtained by binarizing the image by detecting pixels that are equal to or greater than the gradient intensity threshold Pth in the image as edge pixels and classifying the detected edge pixel and other pixels. Is generated (SA9).

  The edge image generation unit 14 compares the gradient directions of the two edge pixels when the thinning processing unit 17 arranges the edge pixels at adjacent pixels in the binarized edge image. When the gradient direction is the same direction and the two edge pixels are aligned in a direction parallel to the gradient direction, the edge pixel having the smaller gradient strength of the two edge pixels is edged. By omitting from the pixels, the binarized edge image is thinned (SA10).

  The feature amount deriving unit 15 is configured to perform a total of four directions including a vertical direction and a horizontal direction of pixels of the binarized edge image finally generated by the edge image generating unit 14 and a 45-degree inclined direction intersecting each other. By calculating the run length matrix respectively, line segment length histograms (continuous edge pixel number histograms) in the four directions are respectively generated (SA11), and the maximum line segment length below the predetermined line segment length threshold Cth is obtained. The average line length from the line segment to the line segment having the line length of the predetermined rank is calculated (SA12), and the discriminant average value that is the average value of the calculated average line segment lengths and The calculated standard deviation, which is the standard deviation of each average line segment length, is derived as a feature amount (SA13).

  When the discriminant average value and discriminant standard deviation as the feature amount are derived by the feature amount deriving unit 15, the scene determining unit 16 captures artifact images taken for various preset buildings. For a population consisting of discriminant average values and discriminant standard deviations as multiple sample values, and for a population consisting of discriminant average values and discriminant standard deviations as multiple sample values in images taken for other than buildings Then, by determining the discriminant average value and discriminant standard deviation as the derived feature amount based on the Mahalanobis's general distance, the artifact stored in the memory 41 is photographed for the building. It is determined whether or not it is an image (SA14).

  The correction processing means 11 performs a correction process based on the determination result for the image to be determined.

  FIG. 13 shows an image determined as an artifact image obtained by photographing a building by the image determination apparatus or the image determination method according to the present invention, and FIG. 14 shows an image determined as an image other than that. It can be seen that an image containing many buildings is determined as an artifact image in which the building is photographed. Further, it can be seen that an image in which natural objects are photographed in most of the images is determined as other images.

  Hereinafter, another embodiment will be described. In the above description, the image conversion unit 12 converts each pixel data of the image to be determined into a value obtained by weighting and adding each of the R component, the G component, and the B component based on [Equation 1]. Thus, the case where the image to be determined is converted into a grayscale image has been described. However, the conversion to a grayscale image is not limited to this, for example, R in each pixel data of the image to be determined It may be converted to a grayscale image by a simple averaging method or a median method that simply averages the component, G component, and B component, or it may be converted to a grayscale image by extracting only a specific color component. Also good. A conversion method according to the property of the image to be determined can be applied as appropriate.

  In the above-described embodiment, the case where the gradient direction Sa (i) is classified and processed in the four directions of the vertical direction, the horizontal direction, and the 45-degree inclined direction intersecting each other has been described. However, the classification processing can be appropriately performed in the direction according to the property of the image to be determined. The classification process is not limited to four directions, and the classification process can be appropriately performed according to the number of directions according to the property of the image to be determined.

  In the above-described embodiment, the case where the binarized edge image is thinned by the thinning processing unit 17 has been described. However, when the determination processing speed is important, the thinning processing can be omitted. .

  The image determination apparatus and image determination method of the present invention can also be applied to simple image classification on a computer, image search on the Internet, and the like.

  In addition, the above-described embodiment is merely an example of the present invention, and it is needless to say that the specific configuration of each block can be changed and designed as appropriate within the scope of the effects of the present invention.

Explanatory drawing of functional block configuration in image processing unit Explanatory drawing of the external configuration of the photographic image processing apparatus Illustration of photo printer Explanatory drawing of functional block configuration of photographic image processing apparatus It is explanatory drawing of a Sobel filter, (a) is a Sobel filter which extracts the vertical direction component of gradient strength, (b) is a Sobel filter which extracts the horizontal direction component of gradient strength. It is explanatory drawing about the production | generation of a binarization edge image, (a) is a gray scale image before binarization edge image generation, (b) is the binarization edge image which binarized the gray scale image. Illustration of gradient strength histogram It is explanatory drawing of a thinning process, (a) is explanatory drawing in case the gradient direction is the vertical direction of the pixel of a binarized edge image, (b) is horizontal direction of the pixel of a binarized edge image. (C) is an explanatory diagram in the case where the gradient direction is a 45-degree tilt direction of one of the pixels of the binarized edge image, and (d) is a binarized edge image in which the gradient direction is the direction. Explanatory drawing when it is the other 45 degree inclination direction of the pixel of It is explanatory drawing of a thinning process, (a) is the binarized edge image before thinning process, (b) is the binarized edge image after thinning process Illustration of thinning process Illustration of line length histogram Flow chart for explaining operation of image scene determination Example of an image judged as an artifact image of a building Example of an image determined as an image other than an artifact image where a building was photographed

Explanation of symbols

1: photographic image processing apparatus 10: input image determination unit 11: correction processing unit 12: image conversion unit 13: gradient strength calculation unit 14: edge image generation unit 15: feature amount derivation unit 16: scene determination unit 17: thinning process means

Claims (14)

An image determination apparatus for determining a scene of an image,
Based on the calculated gradient strength, image conversion means for converting the image to be determined into a grayscale image, gradient strength calculation means for calculating the gradient strength and gradient direction of each pixel of the converted grayscale image, and Edge image generation means for generating an edge image obtained by binarizing an image, and feature quantity derivation means for extracting a line segment composed of a predetermined range of continuous pixels included in the generated edge image and deriving the feature quantity And an image determination apparatus comprising: a scene determination unit that determines an image scene from the derived feature amount.   The image determination apparatus according to claim 1, wherein the gradient intensity calculation unit calculates gradient directions in four directions of a gradient intensity and a vertical direction, a horizontal direction, and a 45-degree inclination direction in pixels of the grayscale image.   The image determination apparatus according to claim 1, wherein the edge image generation unit includes a thinning processing unit that selects a line segment included in the generated edge image based on the gradient direction and the gradient intensity.   The image determination apparatus according to claim 1, wherein the feature quantity deriving unit derives an average length and standard deviation of the extracted line segment as a feature quantity.   The feature amount deriving unit extracts a predetermined number of line segments in order of increasing number of pixels, which is a line segment in which the number of pixels constituting the line segment in each direction is smaller than a predetermined maximum number of pixels. The image determination apparatus according to claim 4, wherein an average of average values in each direction is derived as the average length.   The image determination apparatus according to claim 1, wherein the scene determination unit determines a scene of the image by discriminant analysis based on a Mahalanobis general distance using the feature quantity as a variable.   7. The scene determination unit according to claim 1, wherein the scene determination unit determines whether the image to be determined is a natural image obtained by photographing a natural object or an artificial object image obtained by photographing a building. Image determination device. An image determination method for determining an image scene,
The image conversion step for converting the image to be determined into a grayscale image, the gradient strength calculation step for calculating the gradient strength and gradient direction of each pixel of the converted grayscale image, and the calculated gradient strength based on the above An edge image generation step for generating an edge image obtained by binarizing the image, and a feature amount derivation step for extracting a line segment composed of continuous pixels within a predetermined range included in the generated edge image and deriving the feature amount And a scene determination step for determining a scene of the image from the derived feature amount.   The image determination method according to claim 7, wherein in the gradient intensity calculating step, gradient intensity in pixels of the gray scale image and four gradient directions of a vertical direction, a horizontal direction, and a 45-degree inclined direction are calculated.   The image determination method according to claim 8 or 9, wherein, in the edge image generation step, a thinning process for selecting a line segment included in the generated edge image based on the gradient direction and the gradient intensity is performed.   The image determination method according to claim 8, wherein, in the feature amount derivation step, an average length and a standard deviation of the extracted line segments are derived as feature amounts.   In the feature amount derivation step, a predetermined number of line segments are extracted in order of increasing number of pixels, the number of pixels constituting the line segment in each direction being smaller than a predetermined maximum pixel number, and the extracted line segments The image determination method according to claim 11, wherein an average of average values in each direction is derived as the average length.   13. The image determination method according to claim 8, wherein in the scene determination step, the scene of the image is determined by discriminant analysis based on Mahalanobis general distance using the feature amount as a variable.   The said scene determination step WHEREIN: It determines whether the image used as discrimination | determination object is a natural image by which the natural thing was image | photographed, or an artifact image by which the building was image | photographed. Image determination method.