JP6201549B2 - Image processing apparatus and image processing program - Google Patents

Description

  The present invention relates to an image processing apparatus and an image processing program.

  2. Description of the Related Art Conventionally, there is known an image processing apparatus that determines a target object by comparing a feature amount obtained from an image of the target object with a feature amount of an object candidate stored in advance. As the feature amount, for example, the length of a plurality of radiations drawn at a predetermined angular interval from the reference point of the image of the target object is used.

JP 2012-146108 A

  However, when a “folded portion” is included in the image of the target object, a plurality of intersections between the one radiation and the contour line of the target object image appear for one radiation. May be reduced. As a result, the determination accuracy of the target object may be reduced.

  The disclosed technology has been made in view of the above, and an object thereof is to provide an image processing apparatus and an image processing program capable of improving the determination accuracy of a target object.

In the disclosed aspect, the image processing apparatus includes a processor. The processor calculates the curvature at each of a plurality of division points that divide the contour line of the image of the target object into a plurality of sections . The processor calculates a first histogram representing a frequency with respect to the curvature from a change in the curvature with respect to the permutation of the plurality of division points . The processor calculates a value of a statistical parameter based on the calculated first histogram. The processor identifies a classification to which the target object belongs among a plurality of classifications based on the calculated statistical parameter value. Wherein the processor is based on the correlation of the changes of the curvature, the target object Ru determine Teisu.

  According to the disclosed aspect, it is possible to improve the determination accuracy of the target object.

FIG. 1 is a diagram illustrating an example of an image processing apparatus according to an embodiment. FIG. 2 is a flowchart illustrating an example of a processing operation of the image processing apparatus according to the embodiment. FIG. 3 is a flowchart illustrating an example of a processing operation of the image processing apparatus according to the embodiment. FIG. 4 is a flowchart illustrating an example of the processing operation of the image processing apparatus according to the embodiment. FIG. 5 is a flowchart illustrating an example of the processing operation of the image processing apparatus according to the embodiment. FIG. 6 is a diagram for explaining the extraction of the contour of the image of the target object. FIG. 7 is a diagram for explaining a method of calculating the center of gravity. FIG. 8 is a diagram for explaining the parallel movement and the position vector L [k]. FIG. 9 is a diagram for explaining the division points for dividing the outline into a plurality of sections. FIG. 10 is a diagram for explaining the association process between the dividing point vector and the position vector. FIG. 11 is a diagram illustrating an example of a curvature change curve. FIG. 12 is a diagram illustrating an example of a curvature histogram. FIG. 13 is a diagram illustrating an example of a position vector length histogram. FIG. 14 is a diagram illustrating an example of the classification database. FIG. 15 is a diagram illustrating an example of a classification database. FIG. 16 is a diagram for explaining the classification specifying process to which the target object belongs. FIG. 17 is a diagram for explaining the classification specifying process to which the target object belongs. FIG. 18 is a diagram for explaining the target object determination process. FIG. 19 is a diagram for explaining a specific example of calculating statistical parameter values and determining a target object. FIG. 20 is a diagram for explaining a specific example of calculation of a statistical parameter value and determination of a target object. FIG. 21 is a diagram for explaining a specific example of calculation of a statistical parameter value and determination of a target object.

  Embodiments of an image processing apparatus and an image processing program disclosed in the present application will be described below in detail with reference to the drawings. Note that the image processing apparatus and the image processing program disclosed in the present application are not limited by this embodiment.

[Configuration example of image processing apparatus]
FIG. 1 is a diagram illustrating an example of an image processing apparatus according to an embodiment. FIG. 1 shows a configuration of a mobile phone as an example of an image processing apparatus. In FIG. 1, the image processing apparatus 10 includes an RF (Radio Frequency) circuit 11, a display unit 12, a GPS processing unit 14, a camera 13, a processor 15, and a memory 16.

  The RF circuit 11 performs predetermined wireless transmission processing, that is, digital-analog conversion, up-conversion, and the like on the signal received from the processor 15 to form a wireless signal, and transmits the formed wireless signal via an antenna. Further, the RF circuit 11 performs predetermined reception radio processing, that is, down-conversion, analog-digital conversion, and the like on the reception signal received via the antenna, and outputs the reception signal after the reception radio processing to the processor 15.

  The display unit 12 displays display data received from the processor 15. Further, the display unit 12 has a touch panel function, and outputs information corresponding to the touch position to the processor 15.

  The GPS processing unit 14 acquires the position information of the image processing apparatus 10 and outputs it to the processor 15.

  The camera 13 outputs captured image data to the processor 15.

  The memory 16 stores image data captured by the camera 13, image data acquired via a wireless line, various tables (database), and various programs executed by the processor 15. Examples of the memory 16 include a random access memory (RAM) such as an SDRAM (Synchronous Dynamic Random Access Memory), a read only memory (ROM), and a flash memory.

  The processor 15 executes baseband processing such as encoding, modulation, demodulation, and decoding, and application processing such as image processing. Examples of the processor 15 include a central processing unit (CPU), a digital signal processor (DSP), and a field programmable gate array (FPGA). The processor 15 reads the program from the memory 16 and executes each process. Here, although one processor 15 is assumed to execute baseband processing and application processing, the present invention is not limited to this, and baseband processing and application processing may be executed by different processors. .

  For example, the processor 15 calculates the curvature at each division point that divides the outline of the target object into a plurality of division sections.

  Further, the processor 15 executes a target object determination process based on a curvature change curve with respect to a permutation of a plurality of division points.

  Specifically, the processor 15 calculates a histogram (hereinafter, sometimes referred to as “curvature histogram”) representing the frequency with respect to the curvature from the curvature change curve with respect to the permutation of a plurality of division points. Then, the processor 15 calculates the value of the “statistical parameter” based on the calculated curvature histogram. The statistical parameter calculated from the curvature histogram includes at least one of a curvature radius, a curvature mode value, and a curvature standard deviation. Then, the processor 15 identifies a classification to which the target object belongs among a plurality of “classifications” based on the calculated statistical parameter value. Then, the processor 15 determines, as a target object, an object candidate having the highest correlation with respect to the curvature change curve among a plurality of object candidates belonging to the specified classification.

  The processor 15 may calculate the statistical parameter based on the “position vector length histogram”. The “position vector length histogram” is a histogram of the separation distance from the reference point (for example, the center of gravity) of the target object image to the point on the contour line. The statistical parameter calculated from the position vector length histogram includes the variance value of the separation distance. The image processing by the processor 15 will be described in detail below.

[Operation example of image processing apparatus]
2, 3, 4, and 5 are flowcharts illustrating an example of the processing operation of the image processing apparatus according to the embodiment. FIG. 2 particularly shows an outline of the image processing operation of the image processing apparatus, and FIGS. 3, 4, and 5 particularly show specific examples of the image processing operation of the image processing apparatus. 4 is a flowchart continued from the flowchart of FIG. 3, and FIG. 5 is a flowchart continued from the flowchart of FIG. The flow shown in FIGS. 2, 3, 4, and 5 starts when the processor 15 acquires an image captured by the camera 13 or an image stored in the memory 16.

<Calculation of curvature and curvature histogram>
The processor 15 calculates the curvature at each of the dividing points that divide the contour line of the target object image in the acquired image into a plurality of divided sections (step S101). Then, the processor 15 calculates a curvature histogram from the curvature change curve with respect to the permutation of a plurality of division points (step S102).

  Step S101 and step S102 correspond to step S201 to step S208 in FIGS.

  For example, the processor 15 extracts the coordinates L [N] of the contour of the image of the target object in the acquired image (step S201). Here, N can take each value from 0 to a predetermined natural number n. That is, L [N] is a vector having the origin as the start point and the point on the contour line as the end point. Hereinafter, this vector may be referred to as a “position vector”. For example, when a cat image is acquired, a contour as shown in FIG. 6 is extracted. The cat in the image shown in FIG. 6 has a bent ear, and a portion corresponding to the ear corresponds to the “folded portion”. As the contour extraction method, for example, Hough transform can be used. FIG. 6 is a diagram for explaining the extraction of the contour of the image of the target object.

  Then, the processor 15 extracts the center of gravity G of the contour (step S202). Here, since a plurality of position vectors corresponding to each point on the contour line are obtained in step S201, the center-of-gravity vector G can be calculated using the plurality of position vectors as shown in FIG. FIG. 7 is a diagram for explaining a method of calculating the center of gravity.

  Then, the processor 15 translates the contour extracted in step S201 so that the center of gravity calculated in step S202 overlaps the origin, and creates (calculates) a plurality of position vectors R [k] starting from the origin ( Step S203). That is, a certain position vector R [k] can be obtained by calculating the difference between the position vector L [k] and the centroid vector G. k is an integer from 0 to n. FIG. 8 is a diagram for explaining the parallel movement and the position vector L [k].

Then, the processor 15 normalizes the plurality of position vectors R [k] by dividing each position vector [k] by the length R _Max of the longest one among the plurality of position vectors R [k]. A plurality of position vectors I [k] are calculated (step S204).

  Then, the processor 15 calculates the total length S of the contour line using the plurality of position vectors I [k], and sequentially calculates the accumulated value Sa [k] between the end points of the adjacent position vectors I [k]. Calculate (step S205). As a result, n accumulated values Sa are calculated, so that n numeric strings are obtained. For example, the numerical value sequence of the accumulated values Sa may be thinned out to about half. Let Sb [i] be a numerical sequence of the accumulated value Sa after thinning. Further, as shown in FIG. 9, the processor 15 equally divides the total length S of the contour line by (n−1) and sequentially calculates the cumulative value a [i] between the division points at each division point (step S205). As a result, n accumulated values a [i] are calculated, and thus n numerical strings are obtained. FIG. 9 is a diagram for explaining the division points for dividing the outline into a plurality of sections.

  Then, the processor 15 determines an end point corresponding to the accumulated value Sb whose difference from the accumulated value a at a certain division point (here, the difference between the ratio of the accumulated value Sb to the accumulated value a and 1) is smaller than the predetermined value Δ1. The position vector I is sequentially specified, and the plurality of specified position vectors I are set as a vector string P [j] (step S206). That is, a position vector I close to the line segment length accumulated value a [k] is selected and a correspondence is created to create a vector sequence P [j]. Specifically, as shown in FIG. 10, when there is a dividing point vector ap, aq, ar, as that has the origin as the starting point and the dividing point as the end point, the difference from the accumulated value a corresponding to the dividing point vector. R1, R2, R3, and R4 are specified as the position vector I corresponding to the accumulated value Sb that is smaller than Δ1. That is, the divided vector is approximated to the nearest position vector I. FIG. 10 is a diagram for explaining the association process between the dividing point vector and the position vector.

  Then, the processor 15 calculates the curvature of the contour line at the end point of each P [j], and creates a graph in which the vertical axis is the curvature value and the horizontal axis is the item number j, that is, a curvature change curve (step S207). . FIG. 11 is a diagram illustrating an example of a curvature change curve.

  Then, the processor 15 creates a curvature histogram (see FIG. 12) with the horizontal axis as the curvature and the vertical axis as the frequency from the curvature change curve (step S208). FIG. 12 is a diagram illustrating an example of a curvature histogram.

<Calculation of position vector length histogram>
The processor 15 calculates a “position vector length histogram” (step S103).

  Step S103 corresponds to step S209 in FIGS.

  For example, as shown in FIG. 13, the processor 15 creates a position vector length histogram having the horizontal axis as the normalized length of the position vector I and the vertical axis as the frequency of the length of the position vector I (step S209). ). FIG. 13 is a diagram illustrating an example of a position vector length histogram.

<Calculation of statistical parameter values>
The processor 15 calculates a statistical parameter value based on the calculated curvature histogram and position vector length histogram (step S104).

  Step S104 corresponds to step S210 in FIGS.

  For example, the processor 15 uses the line segment length S / (n−1), the curvature average ΣC [j] / (n−1), the curvature mode C [e] as the statistical parameters (that is, attribute values). The curvature standard deviation σ {C [j]} and the vector length variance σ2 {| P [j] |} are calculated (step S210). The curvature radius, curvature mode value, and curvature standard deviation are calculated from the curvature histogram. Further, the vector length variance, that is, the variance value of the separation distance is calculated from the position vector length histogram.

<Identification of the classification to which the target object belongs>
Based on the calculated statistical parameter value, the processor 15 identifies a classification to which the target object belongs among a plurality of “classifications” (step S105).

  Step S105 corresponds to step S211 and step S212 in FIGS.

  Here, a “classification database” is stored in the memory 16. 14 and 15 are diagrams illustrating an example of the classification database.

  As shown in FIG. 14, classification 1 to classification k are stored in the classification database. And as shown in FIG. 15, each classification | category is matched with the average value (namely, attribute average value) of each statistical parameter in each classification | category. Each classification is associated with members 1 to M. At least one member among the members 1 to M is a representative member. Each member is associated with a curvature change curve (that is, a curvature graph) and a value of each statistical parameter (that is, an attribute value) for each member.

  For example, the processor 15 compares the attribute value calculated for the target object with the attribute average value of each classification to determine whether they are similar, and selects a classification candidate to which the target object belongs (step) S211). For example, the difference between each attribute average value for each classification with respect to each attribute value for the target object (here, the difference between the total ratio of each attribute average value for each classification with respect to each attribute value for the target object and 5) Are sequentially identified that are smaller than the predetermined value Δ2.

  Here, as shown in FIG. 16, it is assumed that classification 1 and classification 3 are selected as classification candidates. 16 and 17 are diagrams for explaining the classification specifying process to which the target object belongs.

  Then, as shown in FIG. 17, the processor 15 compares the attribute value of each representative member of the selected plurality of classification candidates with the attribute value calculated for the target object, and selects the most similar classification (step S212). For example, for each attribute value, the processor 15 sorts the attribute values of a plurality of representative members in the order close to the attribute value for the target object. For example, the processor 15 gives a higher score to a representative member with a higher rank. Then, the processor 15 specifies the classification to which the representative member having the highest total of assigned scores belongs as the classification to which the target object belongs.

<Determination of target object>
The processor 15 determines the target object (step S106). For example, the processor 15 determines (specifies) a candidate object having the highest correlation with respect to the curvature change curve among a plurality of object candidates (that is, members) belonging to the specified classification as a target object.

  Step S106 corresponds to step S213 in FIGS.

  For example, as shown in FIG. 18, the processor 15 calculates the correlation value between the curvature change curve of the image of the target object and the curvature change curve of each member in the specified classification, and targets the member with the highest correlation value. It is determined (specified) as an object (step S213). FIG. 18 is a diagram for explaining the target object determination process.

  Here, the calculation of the value of the statistical parameter and the determination of the target object will be described with a specific example. 19, 20, and 21 are diagrams for explaining specific examples of the calculation of the value of the statistical parameter and the determination of the target object.

  For example, it is assumed that the target object is a wine glass. In the classification database, classifications such as “dogs”, “vehicles”, “glasses”, “plants”,..., “Buildings” are stored.

  In this case, the processor 15 compares the attribute value calculated for the image of the wine glass as the target object with the attribute average value of each classification, and selects a similar classification (that is, a classification candidate). Here, “glass” and “building” are selected as classification candidates.

  Next, the processor 15 determines the attribute value of “Wine Glass” that is the representative member of the classification “Glass”, the attribute value of “Tokyo Tower” that is the representative member of the “Building” class, and the wine glass that is the target object. Are compared with each other, and the classification to which the most similar representative member belongs is specified. Here, the classification “glass” is specified.

  Next, the processor 15 changes the curvature change curve of each of “Jug”, “Wine Glass”, and “Shot Glass” that are members belonging to the classification “Glass”, and the curvature change curve of the image of the wine glass that is the target object. And the member having the largest correlation value is specified. Here, the member “wine glass” is specified.

  The representative member of each classification may be set in advance by the user, or may be set to a member having an attribute value closest to the attribute average value of the classification among the members of the classification. Different representative members may be set for each place. For example, when the position of the image processing apparatus 10 is “Tokyo”, the representative member of the classification “building” may be “Tokyo Tower”, and when it is “Yokohama”, it may be “Yokohama Marine Tower”.

  As described above, according to the present embodiment, the processor 15 in the image processing apparatus 10 calculates the curvature at each of a plurality of division points that divide the contour line of the image of the target object into a plurality of sections. Then, the processor 15 performs a target object determination process based on a curvature change curve with respect to a permutation of a plurality of division points.

  With this configuration of the image processing apparatus 10, curvatures at a plurality of division points set on the contour line of the target object image are calculated. Therefore, even when the target object image includes a “folded portion”, one division is performed. One curvature (that is, one of feature quantities) can be calculated for a point. Accordingly, it is possible to prevent a reduction in the accuracy of the feature amount, and thus it is possible to improve the determination accuracy of the target object.

  Further, in the above determination process, the processor 15 calculates the value of the statistical parameter based on the curvature histogram representing the frequency with respect to the curvature from the curvature change curve with respect to the permutation of the plurality of division points. Then, the processor 15 identifies a classification to which the target object belongs among a plurality of classifications based on the calculated statistical parameter value, and calculates a curvature change curve among the plurality of object candidates belonging to the identified classification. An object candidate having the highest correlation is determined as a target object.

  According to the configuration of the image processing apparatus 10, the classification is roughly specified based on the statistical parameters in the first stage, and the object candidate is refined by applying the curvature change curve to the specified classification in the second stage. It is possible to determine the target object at high speed and with high accuracy.

  Further, in calculating the statistical parameter value, the processor 15 calculates the statistical parameter value based on the position vector length histogram for the separation distance from the reference point of the image of the target object to the point on the contour line. Also good. The statistical parameter calculated based on the curvature histogram includes at least one of a curvature radius, a curvature mode value, and a curvature standard deviation. The statistical parameter calculated based on the position vector length histogram includes the variance value of the separation distance.

  In addition, a representative value related to a statistical parameter, that is, a statistical parameter of a representative member is associated with each of the plurality of classifications. Then, the processor 15 specifies the classification in which the statistical parameter value calculated for the target object is the closest to the representative value as the classification to which the target object belongs.

DESCRIPTION OF SYMBOLS 10 Image processing apparatus 11 RF circuit 12 Display part 13 Camera 14 GPS processing part 15 Processor 16 Memory

Claims (6)

Have a processor,
The processor is
Calculate the curvature at each of a plurality of division points that divide the outline of the image of the target object into a plurality of sections,
Calculating a first histogram representing a frequency with respect to the curvature from a change in the curvature with respect to the permutation of the plurality of division points ;
Based on the calculated first histogram, the value of the statistical parameter is calculated,
Based on the calculated statistical parameter value, the classification to which the target object belongs among a plurality of classifications,
Based on the correlation of the changes of the curvature, Ru determine Teisu the target object in the plurality of object candidate belonging to the specified classification,
An image processing apparatus. The processor is
Calculating a second histogram of the separation distance from a reference point of the image of the target object to a point on the contour line;
In addition to the calculated first histogram, a value of the statistical parameter is calculated based on the calculated second histogram.
The image processing apparatus according to claim 1 . The statistical parameter calculated based on the first histogram includes at least one of a curvature average , a curvature mode value, and a curvature standard deviation.
The image processing apparatus according to claim 1 or 2, characterized in that. The statistical parameter calculated based on the second histogram includes a variance value of the separation distance.
The image processing apparatus according to claim 2 . Each of the plurality of classifications is associated with a representative value related to the statistical parameter,
The processor is
Based on the calculated statistical parameter value and the representative value, the classification to which the target object belongs is specified.
The image processing apparatus according to claim 1 . Calculate the curvature at each of a plurality of division points that divide the outline of the image of the target object into a plurality of sections,
Calculating a first histogram representing a frequency with respect to the curvature from a change in the curvature with respect to the permutation of the plurality of division points ;
Based on the calculated first histogram, the value of the statistical parameter is calculated,
Based on the calculated statistical parameter value, the classification to which the target object belongs among a plurality of classifications,
Determining the target object among a plurality of object candidates belonging to the identified classification based on the correlation with respect to the change in curvature ;
Causing the image processing apparatus to execute processing,
An image processing program characterized by that.

Images