TW548601B - Color image processing method - Google Patents

Abstract

A color image processing method is provided. The color image processing method includes the step of: (a) indexing a color image by assigning representative colors of an image to a color space divided into a plurality of regions. The color image processing method may be applied to object-based image processing such as MPEG-7, and a fast search and retrieval of multimedia contents can be made.

Description

1. Scope of the Invention The present invention relates to a color image processing method, a color image compiling method (hereinafter referred to as editing), for searching a color image. It can be performed efficiently. The invention also includes a color image search method. The database of color image processing methods searches for images of 乂 ==. The image corresponding to the sentence "is appropriate" 2. The related technical description is based on the three directions of the two axes of 0 to 2 5 5 according to the conventional color image processing. Therefore, each of an image is represented. That is, , A storage location, so that the color of the image can be valued at the same time in a matching process. Therefore, a database of custom color capacity is used, and it takes a lot of images. Method, each color is Use a vector in the range from the dimensional color space to represent = pixels are stored in the database with 256 x 256 x 256 color feature vectors. 256 x 256 x 256 color feature vectors must be stored in the database. When searching for a 256 x 256 x 256 image processing method, It takes a lot of time to search the database for an overview of the invention. The above problem, the purpose of the present invention is to provide-a variety of colors. This method can use a small capacity database to compile-a color map and quickly search for and Capture the desired image. Dream Color ::::: The purpose is to provide a medium for storing the program that executes the aforementioned color image processing method. 548601

Λ 5. Description of the invention (2) A further object of the present invention is to provide a color image method, which can be obtained from a color image processing method according to the present invention. In this way, a desired image is searched. In the color image of the cable, another object of the present invention is to provide a program for searching a f-color image. M mistakenly executes the above-mentioned IΊ) "Therefore, in order to achieve the above purpose, the color image processing method of the present invention. This type of color image is provided with an image input area: 2 Α ~ method includes step (a): better Yes, this color image and this ratio is preferably a percentage. • Obtain the typical colors of the image and their ratios. / Step (pa -1) Preferably, this color image processing method 1 represents each typical Colors and their ratios, and (2) Step 2: (b): Store an area identification code in a data area. The sequence number of both is preferably 'the image is divided from multiple domains. For a selected area, the aforementioned color space is preferred in order to achieve the above-mentioned purposes, A = f color and space. Image processing method. This color image processing method provides-one kind of color-feature vector F, & vector The processing method includes steps (aD: taken to represent: F = {{Ci, pm. The first typical color Ci and its ratio, the number of typical colors in the divided region, .j U, where N is an image (A-2): the typical colors '1 疋 彡 丨 an integer between 1 and 1 ^; color center knife' Bottom to color space with lattice structure

Page 6

EH 548601 said ------ Case No. 89106870 Fifth, the description of the invention (3) — cooking and editing: images; and (a ~ 3): the results of editing and editing exist in the form of Wang 1 Gubie In the database. Yu Mu

With the fin two-㈣, the present invention provides a kind of medium, and a characteristic vector f is obtained by using the code. This vector uses the color of the J-th lattice structure _ = gate 2) to assign each typical color to the crystal): Will compile the i:;: ΐ to compile the image; and "-3 are in the database in the form of individual tables. The color image search method provides a search in the database at one level-map Image. = = Color characteristics of the image, search for the typical color of a query image ^ and its ratio = includes step (a): search within the root track; and ^ 7 eighty-eight, in the database of editing The typical color is assigned to the color that is divided into most areas. The color f space of the poor library image is preferably a three-dimensional color space. Inquiry about the typical image and then include step 1 (H): Obtain a predetermined color. Space color = its ", in-points with a majority: the field selects one to be edited. K = obtained = ratio of typical colors, the difference of which is less than Γ, and the pre-threshold in the data is the matching data. Data group 890402701.ptc page 7 548601

Coincidentally, it is better to obtain the same area, image, that is, force 1 f jj this V- few

Sum of the ratios Find the image.

Step (a) must further include step (a-5). • From the confirmation data = the sum of ratio, • and (a-6) • determine a region as the sum of the ratio obtained in the search of the right region and When the difference between the searched images is less than a predetermined threshold, the area is better searched. When any area of the color image searcher is determined to be the searched image, the surrounding areas are performed, and the foregoing steps are performed. The method further includes a step: if there is none, in order to achieve another color feature of the present invention for the adjacency in the color space, including the step (aa ~ 2): at the crystal point, the material is selected to store the difference between the two matching data in one area. Less than the matching data and; and (a'A kind of color image search method, using ^ invention another-layer provides ί) Λ search-image. This color image search method includes the typical colors of a given query area and their ratios; (f + 1 in the three-dimensional color space of the number lattice, select a typical color obtained by == ("); (a_3; j According to the group's corresponding selected crystal point; the data two typical colors and their ratios, and-the area identification code of the image and sequence, as data; (a_4): confirm sentence, -6): determine a region as Search for images, that is, if the ratio of the typical color of the query area in a region ^ :: right data, a predetermined threshold in the sub-selected data group, the data, a-5): get the same from the confirmed data Region ratio =

Amend

548601 SS89106870 V. Description of the invention (5) Between the sum of the percentage and the sum of the ratio of the query area, when the difference threshold is set, the area is the search image. In order to achieve another object of the present invention, the present invention provides a program for storing a color image search method. This medium-1): Obtain the typical colors of a given query area and their ratios. In a three-dimensional color space divided into a majority of lattices, select a typical color obtained by the corresponding step (a-1); (a-3) · · From a data group, corresponding to the selected lattice point; the database makes the typical color and ratio of most data images, and the area identification code of the serial number of an image, as data; (a-4): Indeed, that is, if the ratio of the typical color of the query area in the data set is selected from a predetermined threshold in the selected data group, the data is; (a-5): the total ratio of the same area obtained from the confirmation data): It is determined that a region is a search image, that is, if the difference between the sum of the fraction ratios and the ratio of the query region in a region is a threshold value, the region is a search image. The above objects and advantages of the present invention will be made clearer from the detailed embodiments with reference to the accompanying drawings. Detailed description of the invention Please refer to FIG. 1, which shows a flowchart of a preferred color image processing method according to the present invention. First, confirm the typical colors and their distribution in an image (step 102). Feelings can be expressed as a percentage, and the percentage refers to dividing the number of corresponding pixels by the number of all pixels according to the color classification. The difference is less than a pre-type medium, and includes steps (a; (a ~ 2): Lattice points, the database selects and stores the difference between the data representing the two regions and the matching data 'is less than the matching data and; and (a-6' the obtained hundred is less than a predetermined detail indicating the more consistent embodiment of each color, Here, the typical color assigned should be multiplied by 1

890402701.ptc p. 9 548601

The typical colors identified and their percentages can be represented by a feature vector. That is, when N is the number of typical colors in each area, and i is between ^ and equal to = integer, 'the feature sentence F represented by the i-th typical color Ci and its percentage Pi can be obtained from the following equation Obtain (step 104) · F {{cpp ",, ..., N} T Refer to Fig. 2, which shows a special K-blood obtained from an area of an image; In the image, the feature vector F of the first region R1 thereof is represented by the typical color Cl and its percentage ρ, and the second typical color is represented by the ratio P :, and the third typical color A and its percentage h. Similarly, the feature vectors of other regions R2, R3, and R4 may not be expressed. The ρ-type color is assigned to a three-dimensional lattice-like structure to index (edit) the images and store them in a database in the form of a separate table (step i 0㈠). The results include-area identification, the stopwatch ratio, and the serial numbers of both an image and an area. The main bus and time, ..., please refer to Figure ^ which shows a three-dimensional color two with a lattice-like structure ' This two-dimensional color shirt space is composed of coordinate axes [, lattice-like structure. Each -lattice has -lattice-ΐ, :::, ~, and η6 at its center. The colors belonging to this lattice range are all indexed to these. In the lattice point 0 Now assume that 'the first typical color is included in the lattice having the first lattice point nl'. The first typical color. 2 is included in the crystal having the (k_n lattice point called- In the grid, the third typical color C1 is included in the third crystal

548601

Within the lattice of lattice point n3. Each typical color and its percentage are stored in the data. In the color image processing method according to the invention, this database has a newly defined form. The database is divided into a plurality of storage locations according to categories. The data stored in the storage locations corresponds to the lattice points η, n2, n3 ............. This form of the database individual table, because Data for the typical colors and their percentages of each area are stored separately. FIG. 4 shows a database used for the color image processing method shown in FIG. 1. This database has the form of individual tables. Now, the typical colors and their percentages are stored in the storage position corresponding to the crystal point to which the typical color belongs, together with the area identification marks representing the numbers of both the image and the Xin: area. Then, the data representing the -th typical color q and its percentage 0, together with the area identification code IDt, are stored in positions corresponding to the first lattice point &. = Sample 'will represent the second typical color $. 2 and its percentage data, the bin is in the position corresponding to the (k-1) th lattice point nki. Next, data representing the third typical color q and its percentage A are stored in positions corresponding to the third lattice point ~. Here, the area identification codes iDi, 11 > 2, and 11 > 3 only indicate the same area of the same image. In other words, please refer to the database shown in Fig. 4. The first lattice point h corresponds to the data ID, Ci and. The third lattice point n3 corresponds to the data ID3, c3, and P3, and the U-1) lattice point, the corresponding data%, c2, and p2. According to the above-mentioned fir color image processing method, when an image is edited, the typical colors of the image are assigned to a color space divided into a plurality of regions. So the size of the database used to store the index information of an image is quite small

890402701. ptc page 11 5th 548601 ----- j No. 891Q6870 Description of the invention. With the above-mentioned color image processing method, the efficiency search is similar to a query image. The user wishes to use the image according to the present invention. Please refer to FIG. 5, which shows a flowchart of the image search method. First, it is desired to search in a database. Next, confirm the determined score ratio (step 502). Step 5 The steps 10 and 2 in the method are the same. The typical color will be the value of the pixel obtained by 100. Color images compiled by monthly method can be used for high images. The so-called query image refers to the image searched by the color image search method. According to the preferred embodiment of the present invention, the query area is determined first, that is, the user starts the image. The typical colors in the query area and their 002 and color image processing shown in Figure 1. Here, the percentage refers to the corresponding number divided by the number of all pixels and then multiplied, and then selected the typical color corresponding to each confirmation Step 504): * As shown in FIG. 3, each of the two domains of the lattice is a typical color of the evening number, and it is assumed that in the three-dimensional color space, the t center of each region has a lattice point to distinguish the region. That is, the confirmed typical colors each have a region to which the colors belong in the two-dimensional color space. Therefore, the center point of each region, that is, its lattice point can be selected. In addition, in order to eliminate the wrong cooperation, it is best to consider the typical colors and their distribution at the same time, that is, the percentage of the typical colors. Therefore, a data group is selected from the database t, corresponding to the selected lattice point (step 506), and the matching data is confirmed (step 508). The so-called matching data refers to the typical color percentage of the query area in the data, the difference is less than the selected data

IEMII 890402701.ptc page 12 548601 correction

___Case No. 89106S7n V. Description of the Invention (9) ί sound welding ιΐ ^ I f shows I r 年 year -r I month L: Η 3 Ή, a predetermined threshold in the group board. For example, suppose that a typical color among the typical colors in the query area accounts for about 30% in the query area, and the predetermined value is 5%. Then, if the corresponding lattice point of the data corresponds to the typical color selected in the query area When the same and within the range of 30% ± 5% (30% is selected from the query area, the percentage of the typical color), that is, 25% to 35% of the data, all match the data. For the following ', the sum of the percentages of the same region in the matching data is obtained (step 5 10). In other words, among the data determined as the matching data, the data of the same area is shown by the area identification code, classified, and the total percentage of the data after classification is obtained. Then, in some areas, if the difference between the percentage sum obtained in step 5 10 and the percentage sum of the query area is less than a predetermined threshold value, these areas are determined as searched areas (step 5 1 2). The other two methods are based on the difference in percentage, and the minority area with the smallest difference adjustment can be determined as the searched area, or the single area with the smallest percentage difference can be determined as the searched area. However, according to this method, for example, if the typical color of the query area is located on the outer part of the crystal lattice, although the color of some regions is similar to the typical color, 1 belongs to another lattice. As a result, these areas may not be searched. Therefore, when no area of f is determined as the search area, it is better to perform steps 56 to 512 for the outer peripheral edge lattice points of the lattice points used in the previous search of brothers (step 514). According to the above-mentioned color image search method, the search speed can be accelerated due to the small size of the database; moreover, since image classification is unnecessary

_ Case number 89] flfi87 () (10)

548601 V. Excessive subdivision of the invention description, so it can search efficiently. In addition, the above color search methods can be applied to object-based image processing methods, and achieve fast and efficient multimedia content search and capture. The above color image processing methods and color image search methods can be used in computer programs. In this technical field, skilled computer programmers can easily figure out the codes and code segments that make up a program. At the same time, this program is stored in a computer-readable medium that can be read and executed by a computer, so it can realize the current color image processing method. The aforementioned media include magnetic media, optical media, and carrier waves.

As described above, according to the present invention, the color image processing method can be applied to an object reference image processing method, and can quickly search and retrieve multimedia content.

890402701.ptc page 14

Claims (1)

548601 Case No. 89106870 Amendment VI. Patent Application Scope 11 A color image processing method for indexing an image. This method includes the steps: (a-1): Obtain the i-th typical color c and its ratio p Characteristic direction F = {{ci, pi}, i = 1, ..., N}; where N is the typical number of colors in the segmented area of an image, and i is between 1 and N Integer (a- 2): assign each typical color to each lattice point of the color space with a lattice structure to index the image; and (a- 3): store the index and result in a separate table form The color image processing method in the database 1 2 is the patent application scope No. 1 ^^, wherein the ratio is a percentage. 1 3 The color image processing method according to the first patent application scope No. 1 or No. 1, wherein the marshalling result includes an area identification code, which represents a typical color and its ratio, and an image and an area. Serial number. 14. The color image processing method of the first to the first application of the scope of the patent application, wherein the image is a region selected from a plurality of divided regions. 15. The color image processing method according to claim 1H, wherein the image is a region selected from a plurality of divided regions. 16 The color image processing method according to claim 10 or item 11, wherein the color space is a three-dimensional color space. 17 The color image processing method according to claim 1, wherein the color space is a Sanwei color space. Page 17 548601 _Case No. 89106870 _ Years and months Amendment __ VI. Patent application scope 1 8 The color image processing method as described in the first patent application scope, where the color space is the Sanwei color space. 19 The color image processing method according to the first patent application range, wherein the color space is the Sanwei color space. 2 0—A computer-readable recording medium for storing a program that executes a color image processing method to compile an image, the program includes the steps: (a-1): obtain the i-th typical color c The feature vector F represented by its percentage p: F = {{ci, pi}, i = 1, ..., N}; where N is the number of typical colors in the segmented area of an image, and i is an integer between 1 and N; (a-2): assign each typical color to each lattice point of the three-dimensional color space divided into a plurality of regions to index the image; and (a- 3 ): Store the marshalling results in a database in the form of a separate table. 2 1 Such as the recording medium in the scope of patent application No. 20, where the marshalling results include an area identification code, which represents the typical color and its ratio. , And the sequence numbers of both an image and a region. 22—A color image search method is to search for an image in a database based on the color characteristics of a query image. The method includes step (a): according to the typical colors of a query image and its ratio, The search is performed in the indexed database; the indexing method of the database is to assign the typical colors of the images in the database to the color space divided into most regions. 2 3 The color image search method according to item 22 of the patent application scope, wherein Page 18 548601 _Case No. 89106870_ Year Month Amendment_ VI. Patent Application Scope The color space is a three-dimensional color space. 24. The color image search method according to item 22 of the scope of patent application, wherein step (a) includes the following steps: (a-1): obtaining a typical color of a predetermined query image and its ratio; (a-2) : In a color space divided into a plurality of regions, select a region corresponding to the typical colors obtained in step (a-1); (a- 3): select a data group that is indexed to the selected region from a database Group; and (a-4): confirm matching data, that is, if the ratio of the typical color of the query image in the data is less than a predetermined threshold in the selected data group, the matching data is the matching data. 2 5 The color image search method according to the scope of patent application No. 2 Α_3, wherein step (a) further includes the following steps: (a-5): the sum of the ratios of the same areas obtained from the confirmed data; and (a-6 ): Determine a region as a search image, that is, if the difference between the sum of the ratio obtained and the ratio of the ratio of the query image in a region is less than a predetermined threshold, it is a search image. 26. The color image search method according to the 2U of the patent application scope, wherein the step (a) includes the following steps: (a-1): obtaining the typical colors and ratios of the predetermined query image; (a-2): Within a color space divided into a plurality of regions, a region is selected, corresponding to the typical colors obtained in step (a-1); (a- 3): selecting a data group compiled into the selected region from a database ; And Page 19 548601 _Case No. 89106870_Year_Month__ Sixth, the scope of patent application (a-4): confirm the matching data, that is, if in the data, the difference in the ratio of the typical color of the query image is less than the selected When a predetermined threshold value in the data group is matched data. 27. The color image search method according to the second patent application range, wherein step (a) further includes the following steps: (a-5): the sum of the ratios of the same areas obtained from the confirmed data; and (a-6): It is determined that a region is a search image, that is, if the difference between the total of the obtained ratios and the ratio of the query images in a region is less than a predetermined threshold, it is a search image. 28. The color image search method described in the patent application range 2 ϋ 2 5J, further comprising a step: if no area is determined as a search image, each step is performed in a color space for adjacent peripheral areas 〇29 The color image search method according to any one of the patent application scope No. 2J. No. 2, wherein the ratio is a percentage. 30 The color image search method according to item 2 of the patent application scope, wherein the ratio is a percentage. 3 1 The color image search method according to any one of the patent application scope 2 2_J. 2, wherein the query image is one of the images divided into a plurality of regions, and the search image It is a region in the database image that is divided into a plurality of regions. 32. The color image search method according to the patent application No. 28J, wherein the query image is a region of the image divided into a plurality of regions, and the search image is a database map of the divided region into a plurality of regions Like one of Page 20 548601 _Case No. 89106870_ Year Month Amendment_ VI. Patent Application Area. 33. The color image search method according to the patent application No. 2 9_3, wherein the query image is one of the images divided into a plurality of regions, and the search image is a database image divided into a plurality of regions In a region. 34. The color image search method according to claim 30 of the patent application range, wherein the query image is a region of the image divided into a plurality of regions, and the search image is a database map of the divided region into a plurality of regions An area in the image. 3 5—A color image search method is to search for an image in the database based on the color characteristics of a query image. This method includes the steps: (a-1): Obtain a typical color of a predetermined query area and its ratio. (A-2): In a three-dimensional color space divided into a plurality of lattices, select a lattice point corresponding to the typical color obtained in step (a-1); (a-3): select from a database A data group corresponding to the selected lattice point, and the database stores area identification codes representing the typical colors and ratios of the data images and the serial numbers of both an image and an area as data; And (a-4): confirm the matching data, that is, if the ratio of the typical color of the query area in the data is less than a predetermined threshold value in the selected data group, the matching data. 36. The color image search method according to the patent application No. 3 5 ^ 3, further comprising the following steps: (a-5): the sum of the ratios of the same areas obtained from the confirmed data; and Page 21 548601 _Case No. 89106870_Amendment of the month and year_ VI. Patent application scope (a-6): determine a region as a search image, that is, if the sum of the percentages obtained in a region and the query region When the difference between the sum of the ratios is less than a predetermined threshold, it is a search image. 37. The color image search method according to the patent application No. 36J, further including a step: if no area is determined as the search image, each step is performed in a color space for adjacent peripheral lattice points. 38—A computer-accessible recording medium for storing a program for executing a color image processing method, the program including the following steps: (a-1): obtaining a typical color of a predetermined query area and its ratio; (a -2): In a three-dimensional color space divided into a plurality of lattices, select a lattice point corresponding to the typical color obtained in step (a-1); (a-3): select a data group from a database Group, corresponding to the selected lattice point, and the database stores area identification codes representing the typical colors and ratios of data images and the serial numbers of both an image and an area as data; and (a -4): Confirm the matching data, that is, if the ratio of the typical color of the query area in the data is less than a predetermined threshold in the selected data group, the matching data is the matching data. 39. The recording medium according to item 35 of the scope of patent application, wherein the color image search method further includes the following steps: (a-5): the sum of the ratios of the same areas obtained from the confirmed data; and (a-6) : Determine a region as a search image, that is, if the difference between the total of the obtained percentages and the ratio of the ratio of the query region in a region is less than a predetermined threshold, it is a search image. Page 22 548601 _Case No. 89106870 Sixth, the application for the scope of special thorns to amend the recording media such as the scope of patent application No. 36, where the color image search method further includes a step: if no area is judged as When searching for images, each step is performed on adjacent peripheral lattice points in a color space. Page 23