JP2008112266A - Image processing program and image processing apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processing program and an image processing apparatus that can acquire an optimum filtering pattern in a short time and efficiently. <P>SOLUTION: A mutation part 118 is configured to perform a mutation operation using a 90% or higher mutation rate of genes representing image filter combination patterns if a fitness evaluated by a fitness evaluation part 117 is 50% or lower, or using a 30% or lower mutation rate if the fitness is 80% or higher. Thus, a desired result can be obtained in a short time and efficiently without resulting in a local solution, which is a shortcoming of a genetic algorithm. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an image processing program and an image processing apparatus that determine an optimal combination pattern of image filters to be applied to an image to be processed based on a genetic algorithm, and particularly to obtain an optimal filtering pattern efficiently in a short time. The present invention relates to an image processing program and an image processing apparatus.

  In order to improve the accuracy of image processing such as image recognition, it is important to improve the quality of image filtering processing, which is preprocessing of image processing, in addition to devising the algorithm of the image processing technology itself. The image filtering process is a process of applying an image filter combination pattern to an image to be processed and converting the image into an image suitable for image processing.

  There are many types of image filters, and which one is optimally applied in which order and with which strength depends on the characteristics of the target image and the content of the image processing. Conventionally, since a human determines a combination pattern of image filters (hereinafter simply referred to as “filtering pattern”) based on experience, it has been very difficult to efficiently obtain an optimal filtering pattern.

  In order to efficiently obtain an optimal filtering pattern, Non-Patent Document 1 discloses a technique using a genetic algorithm (hereinafter referred to as “GA: Genetic Algorithm”). Patent Document 1 discloses a technique for increasing the efficiency of GA calculation by increasing the mutation rate in the initial stage of adaptation and lowering the mutation rate as adaptation proceeds.

Masayoshi Maebuchi, Satoshi Ono, Shigeru Nakayama, “Study on Image Filter Design Using Genetic Algorithm”, [online], [October 27, 2006 search], Internet <URL: http: //www.ciec.or. jp / event / 2003 / papers / pdf / E00086.pdf> International Publication No. 2002/061938 Pamphlet

  By the way, in image processing, for example, it may be necessary to obtain a processing result within a short time, such as processing for identifying a license plate of a running vehicle. Like the technique disclosed in Non-Patent Document 1, a method using a general GA method can obtain an optimal filtering pattern in a shorter time than a method of finding a solution with brute force. Still, it took considerable processing time.

  In addition, as in the technique disclosed in Patent Document 1, it is assumed that if the mutation rate is changed according to the fitness, the GA calculation can be speeded up. It was unclear how the mutation rate was set when the fitness was such that the optimal filtering pattern could be obtained most efficiently.

  The present invention has been made to solve the above-described problems caused by the prior art, and an object thereof is to provide an image processing program and an image processing apparatus capable of efficiently obtaining an optimum filtering pattern in a short time. To do.

  In order to solve the above-described problems and achieve the object, one aspect of the present invention is an image processing program that determines an optimal combination pattern of image filters to be applied to an image to be processed based on a genetic algorithm. When the fitness is 50% or less, the mutation rate of the gene representing the image filter combination pattern is at least 90% or more, and when the fitness is 80% or more, the mutation rate is at most 30%. It is characterized as follows.

  In another aspect of the present invention, the image processing apparatus determines an optimal combination pattern of image filters to be applied to an image to be processed based on a genetic algorithm, and when the fitness is 50% or less. The mutation rate of the gene representing the combination pattern of image filters is at least 90% or more, and when the fitness is 80% or more, the mutation rate is at most 30% or less.

  According to another aspect of the present invention, there is provided an image processing method for determining an optimal combination pattern of image filters to be applied to an image to be processed based on a genetic algorithm, when the fitness is 50% or less. The mutation rate of the gene representing the combination pattern of image filters is at least 90% or more, and when the fitness is 80% or more, the mutation rate is at most 30% or less.

  According to another aspect of the present invention, in the above aspect of the present invention, the image having a bit pattern representing presence / absence of use, application order, and intensity of a plurality of image filters is processed based on a genetic algorithm. It is characterized by determining an optimum combination pattern of filters.

  According to the present invention, when the fitness is 50% or less, the mutation rate of the gene representing the image filter combination pattern is at least 90%, and when the fitness is 80% or more, the mutation rate is Since the optimal combination pattern of image filters is determined based on the genetic algorithm at most 30% or less, the desired result can be efficiently obtained in a short time without falling into a local solution that is a weak point of the genetic algorithm. There is an effect that it can be obtained well.

  Exemplary embodiments of an image processing program and an image processing apparatus according to the present invention are explained in detail below with reference to the accompanying drawings.

  First, a general processing flow of image processing will be described. FIG. 1 is a diagram showing a general processing flow of image processing. This figure shows a flow of processing for searching for a portion matching the target image 21 from the original image 11 by pattern matching.

  Prior to the comparison, the original image 11 and the target image 21 are subjected to a filtering process 31 and converted into a processed original image 12 and a processed target image 22, respectively. Then, by executing image processing 32 on the processed original image 12 and the processed target image 22, a portion having the highest degree of coincidence with the target image 21 is selected from the original image 11.

  The filtering process 31 is a process of converting the original image 11 and the target image 21 into an image suitable for the execution of the image process 32 by binarization, smoothing, or the like. A filtering pattern that is a combination of various image filters is converted into an image It is realized by applying to. In order to obtain a good result in the image processing 32, it is necessary to apply an optimal filtering pattern in the filtering processing 31 according to the characteristics of the original image 11 and the target image 21.

  Next, the configuration of the image processing apparatus 100 according to the present embodiment will be described. FIG. 2 is a diagram illustrating the configuration of the image processing apparatus 100 according to the present embodiment. As illustrated in FIG. 1, the image processing apparatus 100 includes a control unit 110, a storage unit 120, an input unit 130, and a display unit 140. The input unit 130 is a device for a user to input instructions and the like, and includes a keyboard and the like. The display unit 140 is a device that displays various graphic information and character information, and includes a liquid crystal display device or the like.

  The control unit 110 is a control unit that controls the entire image processing apparatus 100, and includes an image input unit 111, an initial group generation unit 112, a selection unit 113, a crossover unit 114, a filter unit 115, and an image processing unit 116. A fitness evaluation unit 117 and a mutation unit 118.

  The image input unit 111 is a processing unit that receives input of the original image 11 and the target image 21 and stores them in the image storage unit 121 of the storage unit 120. The initial population generation unit 112 is a processing unit that generates a predetermined number of genes each representing a different filtering pattern and stores the gene in the gene storage unit 122 of the storage unit 120.

  The selection unit 113 is a processing unit that selects a gene pair from the gene storage unit 122, and the crossover unit 114 is a processing unit that performs crossover on the gene pair selected by the selection unit 113. The filter unit 115 applies a filtering pattern corresponding to each gene pair newly generated by the crossover unit 114 to the original image 11 and the target image 21, and sets the pair of the processed original image 12 and the processed target image 22. Two processing units are generated.

  The image processing unit 116 is a processing unit that executes predetermined image processing for each pair of the processed original image 12 and the processed target image 22 generated by the filter unit 115. The fitness evaluation unit 117 is an image processing unit. The processing unit evaluates the fitness for each pair of the processed original image 12 and the processed target image 22 based on the processing result of 116.

  The fitness here is a numerical value indicating how much the processing purpose of the image processing unit 116 is achieved. For example, in the case of pattern matching processing, the processed original image 12 and the processed target image 22 are the best match. Thus, it is obtained based on the magnitude of the difference between the two images (for example, the Euclidean distance) when superimposed. In the case of the pattern matching process, if the fitness is approximately 90% or more, it can be considered that the object is almost achieved.

  The mutation unit 118 is a processing unit that performs a mutation operation on the gene stored in the gene storage unit 122 and changes the content of the gene. The mutation unit 118 dynamically varies the mutation rate according to the higher fitness value obtained by the fitness evaluation unit 117 in order to efficiently obtain the optimum filtering pattern.

  The storage unit 120 is a storage unit that stores various types of information, and includes an image storage unit 121, a gene storage unit 122, and a setting information storage unit 123. The image storage unit 121 is a processing unit that stores the original image 11 and the target image 21 that are image processing targets, and the gene storage unit 122 is a storage unit that stores a gene group representing a filtering pattern. The setting information storage unit 123 is a storage unit in which various setting information for operating the image processing apparatus 100 is stored in advance.

  Next, the processing procedure of the image processing apparatus 100 will be described. FIG. 3 is a flowchart showing a processing procedure of the image processing apparatus 100. As shown in the figure, when the image processing apparatus 100 operates, first, the image input unit 111 receives an input of the target image 21 and stores it in the image storage unit 121 (step S101), and inputs the input of the original image 11. The image is received and stored in the image storage unit 121 (step S102).

  For example, when the content of the image processing is pattern matching processing for finding the target image 21 from the original image 11, the image of the circuit diagram shown in FIG. 4 becomes the original image 11, and the image of the electronic component as shown in FIG. The target image 21 is obtained.

  These images may be input via a medium reading device (not shown) included in the image processing apparatus 100, or transferred from another apparatus through a network to which the image processing apparatus 100 is connected. It may be a thing. In addition, in order to increase the speed of subsequent processing at the stage of storing these images in the image storage unit 121, it is assumed that indispensable filter processing for image processing such as gray scale and smoothing is performed on the image in advance. Also good.

  Subsequently, the initial population generation unit 112 generates genes and stores them in the gene storage unit 122 (step S103). The initial population generation unit 112 generates a predetermined number of bit strings having a predetermined length as genes according to the information stored in the setting information storage unit 123.

  FIG. 6 is a diagram illustrating an example of the meaning of each bit of a gene. In the example shown in the figure, 1 to 4 bits of the gene represent an identifier for identifying the gene, and 5 to 12 bits represent the application order of the filters A to D every 2 bits. The 13 to 20 bits represent the strength at which the filters A to D are applied every 2 bits, and the 21 to 24 bits represent whether the filters A to D are used for each 1 bit.

  As in this example, when a gene has a length of 24 bits, the presence / absence, strength, and application order of the four types of image filters of filters A to D can be stored in each gene. Further, by increasing the number of bits of a gene, information related to a larger number of image filter combinations can be set for each gene, and as in the information for limiting the search area of the target image 21 in the original image 11 Various information can be set for each gene.

  The initial population generation unit 112 may generate a random number by generating a random number, or may generate a gene while adjusting the content bias according to a given setting.

  Subsequently, the selection unit 113 selects two genes from the gene storage unit 122 (step S104), and the crossover unit 114 crosses the two selected genes (step S105). Crossover is a process of exchanging part of two genes. The crossover unit 114 performs one-point crossover, two-point crossover, or the like according to the information stored in the setting information storage unit 123 in order to replace a part of two genes.

  As shown in FIG. 7, the one-point crossing is a method of randomly selecting one point where the gene 41a and the gene 42a cross each other and exchanging the points after that to generate the gene 41b and the gene 42b. As shown in FIG. 8, the two-point crossing is a method of generating two genes 43b and 44b by randomly selecting two points where the genes 43a and 44a cross each other, and exchanging the portion sandwiched between these points. It is.

  Subsequently, the filter unit 115 generates filtering patterns corresponding to the two genes generated by the crossover (step S106), applies the respective filtering patterns to the original image 11 and the target image 21, and the processed original image 12 And two processed target images 22 are generated (step S107).

  Subsequently, the image processing unit 116 performs image processing using the processed original image 12 and the processed target image 22 generated by applying one filtering pattern, and further generated by applying the other filtering pattern. Image processing is executed using the processed original image 12 and the processed target image 22 (step S108).

  Then, the fitness evaluation unit 117 evaluates the fitness of each image processing result (step S109). When the higher fitness is greater than the threshold value stored in the setting information storage unit 123, or the maximum number of generations that have performed a series of mutation processes from selection is stored in the setting information storage unit 123. When the number of generations is exceeded (Yes at step S110), the processing result with the higher fitness is set as the final processing result, and the processing is terminated with the filtering pattern corresponding to the processing result as the optimal filtering pattern.

  Here, even when the number of generations exceeds the preset maximum number of generations, the process is terminated because the fitness does not exceed the threshold even if a series of processes of selection to mutation is repeated. This is to avoid that the process is not completed.

  On the other hand, when none of the termination conditions is satisfied (No at Step S110), the mutation unit 118 determines the mutation rate based on the higher fitness obtained by the fitness evaluation unit 117 (Step S111). Then, the gene corresponding to the mutation rate is selected from the genes stored in the gene storage unit 122, and the contents are changed by performing processing such as partially inverting the bits (step S112).

  After completing the process for one generation in this way, the image processing apparatus 100 returns to step S104 and executes the process for the next generation.

  In the above processing procedure, the mutation unit 118 specifically sets the mutation rate to at least 90% when the fitness is 50% or less, and sets the mutation rate to 80% or more when the fitness is 80% or more. Is at most 30%.

  In the initial stage of searching filtering patterns using GA, at a low fitness level, by setting a high mutation rate, it becomes easier to get close to the optimal solution at an early stage, and GA's weak point that it may fall into a local solution Can be eliminated. In addition, when the fitness level approaches the optimal solution and the fitness level becomes high, it is possible to avoid a rapid decrease in fitness level by setting the mutation rate low.

  FIG. 9 is a diagram showing experimental results for obtaining the optimum value of the mutation rate. The figure shows that the input task is crossed by GA with a fixed mutation rate of 5%, and the fitness at this time is set as the initial fitness, and then the mutation rate is changed according to the fitness. The gene is manipulated to show the fitness in a relatively early generation of the 200th generation.

  The part (a) is the fitness when the process is executed so that the mutation rate is at least 90% or more when the fitness is 50% or less, and the fitness is very high (80% or more). You can see that The portion (b) is the fitness when processing so that the mutation rate is at most 30% when the fitness 80 is greater than or equal to 80%. 90% or more).

  Thus, when the fitness level is 50% or less, the mutation rate is at least 90% or more, and when the fitness level is 80% or more, the mutation rate is at most 30% or less. The filtering pattern having a sufficiently high fitness can be acquired at an early stage.

  FIG. 10 is a diagram showing the transition of the fitness when the mutation rate is fixed, and the fitness does not reach the matching region even if the generations are over 1,500. On the other hand, FIG. 11 is a diagram showing the transition of the fitness when the mutation rate is made variable as described above. The fitness reaches the matching region in a relatively early generation of about 600 generations.

  The configuration of the image processing apparatus 100 according to the present embodiment shown in FIG. 2 can be variously changed without departing from the gist of the present invention. For example, the functions of the control unit 110 of the image processing apparatus 100 are implemented as software, and the functions equivalent to those of the image processing apparatus 100 can be realized by executing the functions with a computer. An example of a computer that executes an image processing program 1071 in which the function of the control unit 110 is implemented as software is shown below.

  FIG. 12 is a functional block diagram illustrating a computer 1000 that executes the image processing program 1071. The computer 1000 includes a CPU (Central Processing Unit) 1010 that executes various arithmetic processes, an input device 1020 that receives input of data from a user, a monitor 1030 that displays various information, and a medium that reads a program from a recording medium. A bus 1080 includes a reading device 1040, a network interface device 1050 that exchanges data with other computers via a network, a RAM (Random Access Memory) 1060 that temporarily stores various information, and a hard disk device 1070. Connected and configured.

  The hard disk device 1070 stores an image processing program 1071 having the same function as that of the control unit 110 shown in FIG. The hard disk device 1070 also includes image data 1072 corresponding to the image stored in the image storage unit 121 shown in FIG. 2, gene data 1073 corresponding to the gene stored in the gene storage unit 122, and setting information storage. Setting data 1074 corresponding to the setting information stored in the unit 123 is stored. Note that the image data 1072, the gene data 1073, and the setting data 1074 may be appropriately distributed and stored in another computer connected via a network.

  Then, the CPU 1010 reads the image processing program 1071 from the hard disk device 1070 and develops it in the RAM 1060, whereby the image processing program 1071 functions as the image processing process 1061. The image processing process 1061 expands information read from the image data 1072 and the like in an area allocated to itself on the RAM 1060 as appropriate, and executes various data processing based on the expanded data and the like.

  The image processing program 1071 is not necessarily stored in the hard disk device 1070. The computer 1000 may read and execute the program stored in a storage medium such as a CD-ROM. . The computer 1000 stores the program in another computer (or server) connected to the computer 1000 via a public line, the Internet, a LAN (Local Area Network), a WAN (Wide Area Network), or the like. You may make it read and run a program from these.

  As described above, in this embodiment, when the fitness evaluated by the fitness evaluation unit 117 is 50% or less, the mutation rate of the gene representing the combination pattern of the image filter is set to at least 90%, When the fitness level of the algorithm is 80% or more, the mutation rate is set to 30% or less, and the mutation unit 118 is configured to perform the mutation operation. The desired result can be obtained efficiently in a short time without falling.

  In the above embodiment, the fitness is calculated by actually executing the image processing on the image to be processed. However, the fitness is calculated by using an evaluation function or the like, and the image processing is performed by optimal filtering. After the pattern is obtained, it may be executed only on the image after applying the optimum filtering pattern.

(Appendix 1) An image processing program for determining an optimal combination pattern of image filters to be applied to an image to be processed based on a genetic algorithm,
When the fitness is 50% or less, the mutation rate of the gene representing the combination pattern of the image filter is at least 90%,
An image processing program characterized in that when the fitness is 80% or more, the mutation rate is at most 30%.

(Additional remark 2) The optimal combination pattern of the said image filter is determined by processing the gene which has a bit pattern showing the presence or absence of a some image filter, an application order, and intensity | strength based on a genetic algorithm, The image processing program according to appendix 1.

(Appendix 3) An image processing apparatus that determines an optimal combination pattern of image filters to be applied to an image to be processed based on a genetic algorithm,
When the fitness is 50% or less, the mutation rate of the gene representing the combination pattern of the image filter is at least 90%,
An image processing apparatus characterized in that when the fitness is 80% or more, the mutation rate is at most 30%.

(Appendix 4) An image processing method for determining an optimal combination pattern of image filters to be applied to an image to be processed based on a genetic algorithm,
When the fitness is 50% or less, the mutation rate of the gene representing the combination pattern of the image filter is at least 90%,
An image processing method characterized in that when the fitness is 80% or more, the mutation rate is at most 30%.

  As described above, the image processing program and the image processing apparatus according to the present invention are useful for determining an optimum combination pattern of image filters to be applied to an image to be image processed, and in particular, shortening an optimum filtering pattern. It is suitable when it is necessary to obtain efficiently in time.

It is a figure which shows the flow of the rough process of an image process. It is a functional block diagram which shows the structure of the image processing apparatus which concerns on a present Example. It is a flowchart which shows the process sequence of an image processing apparatus. It is a figure which shows an example of an original image. It is a figure which shows an example of the target image. It is a figure which shows an example of the meaning of each bit of a gene. It is a figure which shows an example of 1 point crossing. It is a figure which shows an example of 2 point crossing. It is a figure which shows the experimental result for calculating | requiring the optimal value of a mutation rate. It is a figure which shows transition of the fitness at the time of fixing a mutation rate. It is a figure which shows transition of the fitness at the time of making a mutation rate variable. And FIG. 11 is a functional block diagram illustrating a computer that executes an image processing program.

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 Original image 12 Processed original image 21 Target image 22 Processed target image 31 Filtering process 32 Image processing 41a-44a, 41b-44b Gene 100 Image processing apparatus 110 Control part 111 Image input part 112 Initial population generation part 113 Selection part 114 Crossing section 115 Filter section 116 Image processing section 117 Fitness evaluation section 118 Mutation section 120 Storage section 121 Image storage section 122 Gene storage section 123 Setting information storage section 130 Input section 140 Display section 1000 Computer 1010 CPU
1020 Input device 1030 Monitor 1040 Medium reader 1050 Network interface device 1060 RAM
1061 Image processing process 1070 Hard disk device 1071 Image processing program 1072 Image data 1073 Gene data 1074 Setting data 1080 Bus

Claims (3)

An image processing program for determining an optimal combination pattern of image filters to be applied to an image to be processed based on a genetic algorithm,
When the fitness is 50% or less, the mutation rate of the gene representing the combination pattern of the image filter is at least 90%,
An image processing program characterized in that when the fitness is 80% or more, the mutation rate is at most 30%.   2. The optimal combination pattern of the image filters is determined by processing a gene having a bit pattern representing presence / absence of use of plural image filters, application order, and intensity based on a genetic algorithm. The image processing program described in 1. An image processing apparatus for determining an optimal combination pattern of image filters to be applied to an image to be processed based on a genetic algorithm,
When the fitness is 50% or less, the mutation rate of the gene representing the combination pattern of the image filter is at least 90%,
An image processing apparatus characterized in that when the fitness is 80% or more, the mutation rate is at most 30%.