WO2013021889A1 - 色名決定装置、色名決定方法、情報記録媒体、ならびに、プログラム - Google Patents
色名決定装置、色名決定方法、情報記録媒体、ならびに、プログラム Download PDFInfo
- Publication number
- WO2013021889A1 WO2013021889A1 PCT/JP2012/069565 JP2012069565W WO2013021889A1 WO 2013021889 A1 WO2013021889 A1 WO 2013021889A1 JP 2012069565 W JP2012069565 W JP 2012069565W WO 2013021889 A1 WO2013021889 A1 WO 2013021889A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- color
- character string
- image
- document
- search
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/40—Document-oriented image-based pattern recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30176—Document
Definitions
- the present invention relates to a color name determination device, a color name determination method, an information recording medium, and a program, and determines a character string representing the color name of the color from the color.
- HSV Human, Saturation / chroma, Value / brightness
- RGB representation a 24-bit color that uses 3 bytes to represent a color is widely used.
- any numerical value from 0 to 255 is assigned to each element of R, G, and B.
- any numerical value from 0 to 31 is assigned to each element of R, G, and B.
- the accuracy (resolution) of data representation in the color space can be selected as appropriate. Colors expressed with high accuracy can be converted to low accuracy although information is reduced.
- the color name is determined by the standard, the color name that the person actually uses in daily life varies or changes depending on the time and field in which the person is active. In addition, people may name new colors.
- the present invention solves the above-described problems, and is a color name determination device, a color name determination method, an information recording medium suitable for determining a correspondence relationship between a color and a character string representing the color, In addition, it aims to provide a program.
- the color name determination device provides: A color acceptance unit that accepts color input, A keyword extraction unit for extracting a keyword character string related to a document obtained as a result of searching a document in which an image related to the accepted color is arranged; A color appearing in each of a plurality of images obtained as a result of searching for an image related to the extracted keyword character string is expressed with a predetermined accuracy, and the color expressed with the predetermined accuracy is the plurality of images.
- a degree acquisition unit for acquiring the degree of occurrence in each of the A color determination unit that determines a color represented by the received character string from the color expressed with the predetermined accuracy based on the acquired degree;
- a color name determination that determines a keyword character string for the determined color as a color name character string representing the received color It comprises so that a part may be provided.
- the keyword extraction unit may be configured to extract a keyword character string that appears in common in the plurality of searched documents.
- the keyword extraction unit extracts a plurality of the keyword character strings, For each of the extracted keyword character strings, the image search by the image search unit, the degree acquisition by the degree acquisition unit, the color determination by the color determination unit, the color name candidate by the color name determination unit The decision can be configured to be performed.
- the color name determining apparatus of the present invention In the document search, a document described in a language specified by the user who has input the accepted color can be searched.
- the keyword character string can be configured to be a noun, a noun phrase, or a modifier or modifier that modifies the noun in the designated language.
- the color name determination device of the present invention is External image search connected to the color name determination device so as to communicate an image search request including a query specifying a character string meaning color in the specified language and the extracted keyword character string
- An image search unit that transmits to the device, receives an image search response to the image search request from the image search device, and uses a plurality of images specified in the image search response as a result of the image search can do.
- the color name determination device of the present invention is An image search unit for searching for an image related to the extracted keyword character string;
- the degree acquisition unit can be configured to acquire the degree from a plurality of images obtained as a result of an image search by the image search unit.
- the color name determination device of the present invention is A document search request including a query for designating the accepted color is transmitted to an external document search device that is communicably connected to the color name determination device, and a document search response to the document search request from the document search device It is possible to further comprise a document search unit that receives the document specified in the document search response as a result of the document search.
- the color name determination device of the present invention is A document search unit for searching for a document in which an image related to the received color is arranged;
- the keyword extraction unit may be configured to extract the keyword character string based on a document obtained as a result of document search by the document search unit.
- the color name determination method includes: A color acceptance process for accepting color input; A keyword extraction step of extracting a keyword character string related to a document obtained as a result of searching a document in which an image related to the accepted color is arranged; A color appearing in each of the plurality of images obtained as a result of searching for an image related to the extracted keyword character string is expressed with a predetermined accuracy, and the color expressed with the predetermined accuracy is the plurality of colors A degree acquisition step of acquiring the degree of appearance in each of the images; A color determining step for determining a color represented by the received character string from the color expressed with the predetermined accuracy based on the acquired degree; When the received color and the determined color are the same or similar, a color name determination that determines a keyword character string for the determined color as a color name character string representing the received color It comprises so that a process may be provided.
- a computer-readable information recording medium is provided.
- Computer A color acceptance unit that accepts color input, A keyword extraction unit for extracting a keyword character string related to a document obtained as a result of searching a document in which an image related to the accepted color is arranged; A color appearing in each of a plurality of images obtained as a result of searching for an image related to the extracted keyword character string is expressed with a predetermined accuracy, and the color expressed with the predetermined accuracy is the plurality of images.
- a degree acquisition unit for acquiring the degree of occurrence in each of the A color determination unit that determines a color represented by the received character string from the color expressed with the predetermined accuracy based on the acquired degree;
- a color name determination that determines a keyword character string for the determined color as a color name character string representing the received color It is configured to record a program that functions as a part.
- the program according to the fourth aspect of the present invention is: Computer A color acceptance unit that accepts color input, A keyword extraction unit for extracting a keyword character string related to a document obtained as a result of searching a document in which an image related to the accepted color is arranged; A color appearing in each of a plurality of images obtained as a result of searching for an image related to the extracted keyword character string is expressed with a predetermined accuracy, and the color expressed with the predetermined accuracy is the plurality of images.
- a degree acquisition unit for acquiring the degree of occurrence in each of the A color determination unit that determines a color represented by the received character string from the color expressed with the predetermined accuracy based on the acquired degree;
- a color name determination that determines a keyword character string for the determined color as a color name character string representing the received color It is configured to function as a part.
- the program of the present invention may be recorded on a computer-readable non-transitory information recording medium such as a compact disk, a flexible disk, a hard disk, a magneto-optical disk, a digital video disk, a magnetic tape, and a semiconductor memory. It can.
- This information recording medium can be distributed and sold independently of the computer.
- the above program can be distributed and sold via a transmission medium such as a computer communication network, independently of the computer on which the program is executed.
- a color name determination device it is possible to provide a color name determination device, a color name determination method, an information recording medium, and a program suitable for determining the correspondence between a color and a character string representing that color. .
- RGB Red, Green, Blue
- CYMK CYMK
- HSV H-semiconductor
- each color expresses a color using three numerical values of R, G, and B. That is, each color is expressed as a point or a position vector arranged in a three-dimensional color space composed of the R axis, the G axis, and the B axis. HSV is expressed in a three-dimensional color space, but CYMK has a four-dimensional color space.
- the sameness or similarity between a certain color and another color can be determined by the distance in the color space. That is, the distance between the two colors is determined by the sum of the absolute values of the differences between the components (Manhattan distance) or the square root of the square sum of the differences between the components (Euclidean distance). If the colors are the same color and this distance is smaller than a certain threshold, it is determined that the two colors are similar.
- a user In the application technology of document search, a user generally accesses a search engine through a browser or a dedicated application and gives a query specifying a character string to be searched. Then, the search engine searches the entire web, the LAN, or the computer, and returns a document related to the character string as a search result.
- Some image search techniques search for images using a character string as a query. This is based on an empirical rule that in a document in which an image is arranged, a character string describing the image often exists in the vicinity of the image. For this reason, a search technique similar to the above document search can be used.
- an image search technology that, when a color is specified, searches for an image containing that color.
- a search target image one or a predetermined number of representative colors of the image is selected from the color distribution of pixels included in the image. Then, by determining the degree of similarity between the color designated by the user and the representative color of the image as described above, an image related to the desired color can be obtained as a search result.
- search techniques often employ various similarities, calculate a score indicating how well they match the query for each search result, and present the search result in descending order of the score. Therefore, it is common to use only a predetermined number of items having higher scores instead of using all search results that satisfy the query.
- a color determining device that, when given a character string, determines the color represented by the character string;
- a color name determination device that determines a character string (name of the color, color name) expressing the color;
- An example is given and demonstrated in the order of the example which applied said (1) and (2) to the color of goods.
- one color may be determined, or a plurality of candidate colors may be determined.
- the character string of the color name to be determined may be one, or a plurality of character strings may be color name candidates.
- color determination devices and color name determination devices are typically realized by executing predetermined programs on various computers such as personal computers and web server computers.
- a computer is a CPU (Central Processing Unit) that executes a program, and uses RAM (Random Access Memory) as a temporary storage area or output of processing results.
- Hardware that accepts an instruction from a user by a device, outputs a processing result to an output device such as a display, or performs the above input / output by communicating with other devices via a NIC (Network Interface Card)
- NIC Network Interface Card
- the present embodiment relates to a color determination technique for determining a color represented by a character string from a character string considered by the user to represent the color.
- a character string of length 1 what consists of only one character (for example, "red” “yellow” “green” etc.) is handled as a character string of length 1.
- FIG. 1 is an explanatory diagram showing a schematic configuration of a color determination device according to the present embodiment
- FIG. 2 is a flowchart showing a flow of color determination processing executed by the color determination device.
- the color determination apparatus 101 includes a character string reception unit 102, an image search unit 103, a degree acquisition unit 104, and a color determination unit 105.
- each of these units is typically realized by executing a program on a computer.
- the character string receiving unit 102 receives an input of a character string (step S151).
- the character string received by the character string receiving unit 102 is input by a user or another device (in the color name determination device described later, a specific element of the device gives the character string receiving unit 102 a character string).
- a character string which is assumed to be a character string representing a color name representing a color.
- Various character strings can be used as color names depending on the region, industry, and time, and manufacturers and stores can select color names themselves or create new ones by coined words. Sometimes.
- the character string accepting unit 102 is configured to accept an arbitrary character string.
- the image search unit 103 searches for an image related to the accepted character string, and obtains a plurality of images as a result of the image search (step S152).
- the image search unit 103 uses the image search technique described above to search for an image related to the accepted character string.
- an image to be searched is published on the Internet, is arranged in a web document, and includes a character string accepted in the web document. .
- the image search unit 103 can also use an external image search device, as will be described later. Further, when a computer that implements the color determination apparatus 101 including the image search unit 103 functions as an image search engine, the image search unit 103 does not need to communicate with an external computer.
- any image search technique can be applied.
- the degree acquisition unit 104 expresses the color appearing in each of the plurality of images with a predetermined accuracy, and acquires the degree of appearance of the color expressed with the predetermined accuracy in each of the plurality of images (step S153). ).
- the degree acquisition unit 104 acquires the degree of color representing what color and how much color is included in the image obtained as a result of the image search.
- the degree of color may be too small. Therefore, typically, the accuracy of the color is lowered, and similar colors are collectively obtained to obtain the degree.
- the color accuracy is lowered and the color space is re-divided into a smaller number of sections.
- a technique for collecting similar values by lowering the accuracy of colors there has been proposed a technique for collecting similar values by lowering the accuracy of colors, and each section obtained as a result is called a “bucket”.
- the number of divisions in the color space is 32 ⁇ 32 ⁇ 32 (about 30,000), and if the color accuracy is set to 12-bit color, the number of divisions is 16 ⁇ 16. ⁇ 16 pieces (about 4,000 pieces). With this number of colors, the degree of color can be acquired appropriately.
- the degree of a certain color is obtained after the color accuracy is lowered.
- the following can be used.
- (1) The number of images in which the color appears in the search result images.
- (2) The total number of pixels of the color among the pixels of the search result image.
- (3) A value obtained by combining (1) and (2) above. That is, the value is determined so as to increase as the number of images in which the color appears is larger and as the total number of pixels of the color is larger.
- the value in any one of (1), (2), and (3) is equal to or greater than a predetermined threshold, the value or 1 is the degree, and when the value is less than the predetermined threshold, 0 is the degree. .
- the color determination unit 105 determines the color represented by the accepted character string from the colors expressed with a predetermined accuracy based on the acquired degree, and ends the color determination processing (step) S154).
- the color represented by the accepted character string is determined.
- a method for determining only one color the following can be used.
- the color having the highest degree is set as a color represented by a character string.
- the color farthest from the origin in the color space is set as the color represented by the character string.
- the RGB expression is adopted, the brightest color is selected.
- the average color of the largest degree is set as a color represented by a character string.
- a predetermined number of colors having higher degrees are selected, and the average color is set as a color represented by a character string.
- Techniques for determining one or more colors include (1) The color having the highest degree is set as a color represented by a character string. (2) The upper predetermined number of colors having a high degree is set as a color represented by a character string.
- the color determination apparatus 101 applies the image search technology to the character string received from the user or the like, and appropriately selects the color represented by the character string from the colors included in the obtained image. Can be determined.
- the number of images including the color among the search result images is adopted as the degree of color in the first embodiment.
- RGB color space divided into 10 ⁇ 10 ⁇ 10 sections is adopted, “green color” is given as a character string representing a color, and an image search is performed on a document in the Internet.
- 3A, 3B, 3C, and 3D are explanatory diagrams showing the distribution of colors appearing in the search result image.
- a description will be given with reference to FIG.
- the origin of the color space is located below the center, the R axis (Red) is directed to the right, the G axis (Green) is oriented upward to the left, and the B axis (Blue) is oriented upward.
- the position at is marked with a circle.
- FIG. 3A shows the distribution of colors that appear in the top one image of the search results.
- FIG. 3B shows the distribution of colors that appear in common in the top two images of the search results.
- FIG. 3C shows the distribution of colors that appear in common in all the top three images of the search results.
- FIG. 3D shows the distribution of colors that appear in common in all the top four images of the search results.
- the color that appears in common with the upper image of the search result can be found.
- This color is considered to be the color for a given character string.
- the number of colors appearing in common is checked and when it becomes 1, the repetition is stopped there. Then, the remaining one color may be determined as the color for the given character string.
- the search results of the image search are typically arranged in the order of score with the highest degree of matching of the given character string. Therefore, when the search result is N images, the distribution of colors appearing in the top 1 image is obtained, the distribution of colors appearing in all the top 2 images is obtained, and the top 3 images are obtained. If the distribution of colors appearing in all is obtained and the process is repeated as in..., It is considered that the convergence proceeds at high speed.
- N images as search results and a color having a degree of N that is, when a plurality of colors appearing in all N images are obtained, all of them are given as described above.
- it may be determined as a color candidate for a character string, or an average color thereof may be determined as a color for a given character string.
- FIG. 4 is a flowchart showing a control flow of the color determination process in the present embodiment.
- a description will be given with reference to FIG. In this example, only one color for a given character string is determined.
- step S151 When the color determination process according to the present embodiment is started, the character string reception (step S151) and the image search (step S152) are executed as in the first embodiment. Thereafter, the following processing is executed.
- the CPU stores, in the RAM, a three-dimensional bit array F [-,-,-] in which one bit is assigned to each section of the color space divided with a predetermined accuracy, and a three-dimensional bit array G [-,-, -] Is secured (step S161).
- the bit array F [-,-,-], G [-,- ,-] Takes a value between [0,0,0] and [M-1, M-1, M-1].
- ⁇ means substitution.
- “*” is used to indicate that processing is performed for all subscripts within the range (the same applies hereinafter).
- the CPU repeats the following process for each pixel included in the image whose search result score is r-th (step S165).
- the CPU acquires the color of the pixel (step S166).
- the color of the pixel is expressed as (r, g, b) in 24-bit color. That is, r, g, and b all take any integer value from 0 to 255.
- multiplication “ ⁇ ” is executed in preference to division “/”, division is integer division, and the remainder is rounded down.
- This calculation means that the color (r, g, b) of the pixel in the image has been converted to a color (i, j, k) with a predetermined accuracy.
- the above conversion formula may be appropriately changed according to the accuracy.
- the CPU converts the bit values in the bit array G [-,-,-] to G [i, j, k] ⁇ G [i, j, k] or 1 (Step S168).
- “or” means bit OR operation.
- “and” means a bitwise AND operation.
- the CPU extracts a subscript having a value of 1 from the bit array G [-,-,-] (this subscript is a numeric string associated with each partition) (step S171).
- the CPU checks the number of extracted subscripts (step S172).
- step S172; 1
- the CPU selects 24-bit color for the section (i, j, k) represented by the extracted subscript [i, j, k].
- the color (r, g, b) is obtained (step S173).
- r (i ⁇ 255) / M;
- g (j ⁇ 255) / M;
- b (k ⁇ 255) / M Should be calculated.
- the CPU outputs the color (r, g, b) as the determined color (step S164), and ends this process.
- step S172; 0
- the CPU extracts a subscript having a value of 1 from the bit array F [ ⁇ , ⁇ , ⁇ ] (step S175).
- L subscripts having a value of 1 in the bit array F [ ⁇ , ⁇ , ⁇ ] as follows. [i [0], j [0], k [0]], [i [1], j [1], k [1]], ..., [i [L-1], j [L- 1], k [L-1]]
- the CPU obtains an average color (r, g, b) of 24-bit color for each section expressed by the extracted subscript (step S176).
- the CPU outputs the color (r, g, b) as the determined color (step S177), and ends this process.
- step S172 the CPU copies each bit of the bit array G [-,-,-] to the bit array F [-,-,-] (step S178).
- this repetition it means that the color could not be narrowed down to one, so the CPU advances the control to step S165.
- the processing speed can be increased and the consumption memory can be suppressed.
- the frequency at which the color appears in the search result image is adopted as the degree of color in the first embodiment.
- the degree acquisition unit 104 obtains a partition in the color space by converting the color of the pixel to a predetermined accuracy for all pixels included in the N images of the search results. Then, for each section, the number of pixels having colors that belong to the section is counted.
- the total number of pixels assigned to each section corresponds to the degree to the color (color with a predetermined accuracy) of the section.
- RGB color space divided into 10 ⁇ 10 ⁇ 10 sections is adopted, “green color” is given as a character string representing a color, and a document in the Internet is searched for an image. .
- FIG. 5 is an explanatory diagram showing the frequency of colors appearing in the search result image.
- FIG. 5 is an explanatory diagram showing the frequency of colors appearing in the search result image.
- This figure shows the frequency distribution of pixel colors in the same image as in FIG. 3A.
- the size of the circles arranged in the color space is changed according to the number of pixels allocated to each section. Yes.
- the color represented by the character string can also be determined by setting the appearance frequency of the color to a degree.
- the color determination unit 105 selects a section having the maximum total number of classified pixels from the color space. Then, the color of the section is output as a color expressed by a character string.
- the average color of these colors may be output as a color represented by a character string.
- the average calculation method may be a simple average, or may be a weighted average using the number of pixels in each section as the color weight of the section.
- the color represented by the character string can be appropriately determined based on the appearance frequency of the color of the pixel appearing in the image.
- the appearance frequency is set to a degree as in this embodiment, it is considered that the number of images for which the color frequency must be counted among the images obtained as a search result can be reduced.
- the degree of color is determined for about 100 images
- the color of about 10 to 20 images is determined. It is expected that sufficient results can be obtained by performing processing for obtaining the degree (frequency of appearance).
- weighting is performed with the frequency in the present embodiment. It is a technique.
- the threshold number for narrowing down candidates is a constant (for example, 10 or 20, etc.), a predetermined number for the total number of sections in the color space (0.05, 0.1, etc., a value exceeding 0 and less than 1).
- the value obtained by multiplying can be adopted.
- a predetermined number (0.05, 0.1, etc.) is added to the number of colors (the number of sections in which the color is distributed in the color space) included in the top first to predetermined positions of the search results.
- a value obtained by multiplying by 0 to a value less than 1 may be used as the narrowing threshold number.
- This embodiment is a modification for removing useless colors from the search result image in the above embodiment.
- the colors of all the pixels included in the search result image are targeted for processing for calculating the degree.
- the color represented by the character string can be determined more accurately by removing the background color arranged around from the target of the degree calculation.
- the degree acquisition unit 104 performs the following preprocessing.
- the pixels having the same color as any of the four corners are removed from the outermost peripheral pixels of the image.
- the image after removal is not necessarily rectangular.
- the pixel adjacent to the removed pixel becomes the outermost pixel.
- the pixels having the same color as any of the colors at the four corners are removed from the pixels on the outer periphery.
- This process is repeated, and when there are no pixels of the same color as the four corner colors in the outer peripheral pixels, this pre-processing is terminated.
- the objects to be removed from the periphery of the image may include not only pixels having the same color as any of the colors at the four corners but also pixels having colors similar to any of the colors at the four corners.
- the pixels at the four corners may be used instead of the pixels at the four corners, for example, only the upper left pixel.
- the background pixels arranged in the peripheral portion of the image can be removed from the processing target, and the pixel having the color of interest is prioritized and becomes the processing target.
- the color represented by the string can be determined.
- This embodiment is a mode suitable for the case where the image search is executed by an external image search apparatus.
- an image search apparatus equipped with image search technology gives a character string as a query, it returns an image in the Internet related to the character string as a search result.
- Such an image search apparatus is communicably connected to a computer that realizes the color determination apparatus 101 via a NIC.
- the image search unit 103 transmits an image search request including a query for designating a character string to the external image search device via the NIC.
- the image search apparatus that has received the image search request arranges the images that match the query among the images published on the Internet in the order of the score of the suitability, and sets it as the search result.
- an image search response designating the search result is transmitted to the color determination apparatus 101.
- the image search unit 103 of the color determination device 101 receives this image search response and gives the image specified in the image search response to the degree acquisition unit 104 as a result of the image search.
- the name of the object having the color may be adopted as the character string representing the color.
- a character string representing a color is often a part of a proper noun such as the name of a person or the name of an organization.
- an image that is not necessarily appropriate may result in a search result.
- the language currently used by the user is used as it is, the character code of the character included in the character string is estimated, the user is directly designated, etc.
- the language currently used by the user is used as it is, the character code of the character included in the character string is estimated, the user is directly designated, etc.
- Various embodiments are possible.
- the language of the character string is estimated by checking which country's character code is assigned to the character code. Can do.
- a character string is input via a browser, the language of the input form currently displayed in the browser is used, or what language is used as the default language or font in the browser.
- the language of the character string may be estimated from the information whether it is.
- the English word “pink” may also be used as a part of a celebrity's group name or stage name, just like “Midori”.
- the character strings “color”, “color”, and the like meaning color are added to the query, so that the color for the character string can be determined more accurately. Will be able to.
- the color represented by the character string is determined to obtain the correspondence between the character string and the color.
- the character string representing the color is determined to determine the correspondence between the character string and the color, and the reverse function of the color determination device 101 is obtained. .
- FIG. 6 is an explanatory diagram showing a schematic configuration of the color name determining apparatus according to the present embodiment.
- FIG. 7 is a flowchart showing the flow of color name determination processing executed by the color name determination apparatus.
- the color name determination device 301 includes a color reception unit 302, a document search unit 303, a keyword extraction unit 304, a color name determination unit 305, and the image search unit 103 in the color determination device 101 described above.
- An acquisition unit 104 and a color determination unit 105 are included.
- each of these units is typically realized by executing a program on a computer.
- the color reception unit 302 receives a color input (step S351).
- the color reception by the color reception unit 302 is performed in the following manner, for example.
- a user directly inputs numerical values of color components in a color space, such as RGB values and HSV values.
- the user designates a specific pixel in the image. Then, the color of the pixel is accepted. Any image can be used, but an image of a color sample in which various colors are drawn with gradation may be adopted.
- the representative color of the image is accepted. For example, after the background pixels of the designated image are removed by the technique of the fourth embodiment, the average color of the remaining pixels is used as the representative color.
- the document search unit 303 searches for a document in which an image related to the accepted color is arranged (step S352).
- image search technology includes image search technology that, when a color is specified, searches for an image including that color.
- the image search apparatus that realizes such an image search based on color is realized by the computer itself that implements the color name determination apparatus 301 executing a program in the same manner as the image search apparatus that realizes an image search based on a character string. It may be realized, or may be realized by another computer communicatively connected via the NIC.
- the keyword extraction unit 304 extracts a keyword character string related to the searched document (step S353).
- the keyword string is the simplest to extract from the character string that appears in the searched document, but comments, tags, and explanations attached to the document or images placed in the document. For example, information that is not directly included in the retrieved document, but is associated with the document, may be extracted from the keyword character string.
- a keyword character string it is desirable to extract words or word strings such as nouns, noun phrases, adjectives, adjective phrases, etc., for example, in Japanese, a character string representing a word and a word that modifies the word.
- the text of the retrieved document can be divided into words, and each word can be adopted as a keyword character string.
- a character string appearing in a document a character string that can be directly viewed by a viewer when a searched document is browsed can be adopted, but an annotation or tag attached to an image arranged in the document. , Comments, alt attribute character strings, and the like that may not be directly visible to the viewer.
- the color name determination device 301 repeats the following processing for each of the extracted keyword character strings (step S354).
- the image search unit 103 searches for an image related to the keyword character string, obtains a plurality of images as a result of the image search, and the degree acquisition unit 104 selects colors that appear in each of the plurality of images. Representing with a predetermined accuracy, the degree of appearance of the color expressed with the predetermined accuracy in each of the plurality of images is acquired, and the color determination unit 105 is expressed with a predetermined accuracy based on the acquired degree.
- the color represented by the accepted character string is determined from the selected color (step S355).
- the color represented by the keyword character string is determined for the keyword character string to be processed in the current iteration.
- the color name determination unit 305 determines whether the determined color (determined color) and the received color (input color) are the same or similar (step S356).
- Identical and similar determination criteria for colors can be determined by the color distance in the color space as described above.
- step S356 If they are the same or similar (step S356; Yes), the keyword character string to be processed in the current iteration is determined and output as a candidate for a color name character string representing the accepted color (step S357). ), And the process after step S354 is repeated (step S358).
- step S356 if they are not the same or similar (step S356; No), the processing after step S354 is repeated as it is (step S358).
- the keyword character string can be considered to express the target color.
- the keyword extracted from the search result document is considered to be a common word and does not represent the target color. It is done.
- step S354 to step S357) for each of the keyword character strings thus extracted is completed, this processing is terminated.
- the keyword extraction unit 304 extracts one or more keyword character strings, and the image search by the image search unit 103 and the degree acquisition unit 104 for each of the extracted keyword character strings.
- the acquisition of the degree of the color name the color determination unit 105 determines the color
- the color name determination unit 305 determines the color name candidate.
- the color name determination device 301 applies the image search technique to the color received from the user or the like, and gives each keyword character string included in the obtained document to the color determination device 101. Based on the similarity between the obtained color and the color received from the user or the like, an appropriate color name representing the color received from the user or the like can be determined from the keyword character string.
- a plurality of keyword character strings are output as character string candidates representing colors.
- This embodiment relates to a preferred embodiment for narrowing down the keyword character string.
- the document search unit 303 is configured to use a plurality of documents as search results.
- a plurality of search results are obtained in the order of suitability for a query. Therefore, in this embodiment, a predetermined number of documents having higher suitability for a given color are used as search results. To get.
- the keyword extraction unit 304 extracts keyword character strings that appear in common in a plurality of searched documents.
- Keyword character string candidates For example, a linear sum of the frequency and the number by a predetermined coefficient can be used as the evaluation value.
- the keyword character string is processed in descending order of the appearance frequency, the number of appearing documents, the evaluation value, etc., and the predetermined upper limit number is set. If the character string representing the color is determined and output as a candidate, the repetition may be aborted.
- keyword character strings are extracted in the order in which it is expected that the target color is highly likely to be expressed, so that character string candidates of a desired number of color names can be quickly obtained. Can be determined.
- the language of the character string representing the color is limited in the color name determining apparatus 301 described above.
- the color determination device 101 uses the character code of the accepted character string, the language of the input form displayed on the browser to accept the character string, and the language used in the default language and default font in the browser. Had acquired the language used by.
- the user explicitly specifies the language, or the language of the input form displayed in the browser, the language used in the default language or the default font in the browser,
- the language specified implicitly by the user is the language desired by the user.
- the document search unit 303 searches for a document described in a language specified by the user.
- the extracted keyword character string is also the user-desired language
- the character string determined to represent the color name is also the user-desired language.
- the information obtained as the language desired by the user here can also be used in the color determination process. That is, the image search unit 103 adds a character string (“color”, “color”, etc.) in a user-desired language meaning the concept of color to the query.
- the character string representing the color can be obtained in the language desired by the user. Therefore, it becomes possible to obtain a color name peculiar to the language culture, and to deal with circumstances peculiar to the country, the language, and the culture.
- the color determination apparatus 101 determines a method for determining only one color represented by the color name, a method for determining one or more, Is disclosed (see the description relating to step S154).
- the upper predetermined number of colors having a high degree is simply selected.
- a clustering technique is applied to more precisely determine a plurality of candidates as colors represented by a character string, and an index indicating the degree of appropriateness that each candidate is represented by the character string The value can be output.
- the color of the pixels included in the search result image is classified into each section (bucket), and the number of pixels belonging to each section is totaled.
- a plurality of representative colors are selected by performing a clustering process on the color distribution in the three-dimensional color space. This is a process corresponding to so-called “color reduction”. Accordingly, one or more appropriate representative colors are selected by various color reduction methods in computer graphics.
- K-means A technique known as K-means is widely known as a technique applicable to color reduction.
- K-means (A) pixel color distribution; (B) a positive integer K which is the number of divisions; Is given, (X) K representative colors; (Y) the strength of the cluster (the number of pixels belonging to the cluster represented by the representative color, the ratio of the number of pixels classified into the cluster to the total number of classified pixels, etc.); Is output.
- K-means a positive integer K which is the number of divisions; Is given, (X) K representative colors;
- Y the strength of the cluster (the number of pixels belonging to the cluster represented by the representative color, the ratio of the number of pixels classified into the cluster to the total number of classified pixels, etc.); Is output.
- the number of colors represented by a certain color name is considered to be appropriate up to about three.
- clustering may be performed by K-means by specifying a value of about 3 to 6 as K.
- cluster strength can be assigned as an index of “appropriateness” in which the representative color is expressed by a color name.
- the clustering calculation may be speeded up, and the data may be truncated to remove the color corresponding to the noise. For example, when there are 100 original images, only a section having 40 or more pixels belonging to the section is subjected to clustering, and a section of less than 40 is ignored.
- the threshold value for ignoring can be determined based on the number M of original images and the total number G of original pixels with respect to the total number L of sections. For example, if the number of pixels belonging to a partition is less than M / (A ⁇ L) using appropriate positive constants A and B, it is ignored, or the number of pixels belonging to the partition is G / (B ⁇ Ignore if less than L), etc. This threshold value may be appropriately designated by the user.
- a color determination device 101 that outputs one or a plurality of colors can be prepared when a character string of one color name is given, this can also be applied to the color name determination device 301.
- step S355 for each character string extracted in step S353 of the sixth embodiment, one or more colors (determined colors) are determined in step S355.
- step S356 it is determined whether each of the determined colors is the same as or similar to the color (input color) received from the user in step S351. If they are similar, the character string that is the source of the determined color may be output as a candidate for calling the input color.
- clustering technology applicable to this embodiment is not limited to K-means or X-means, and even when various clustering technologies are used, they are included in the technical scope of this embodiment.
- the character string received by the color determination device 101 is a character string describing the color of the product, and is selected by the manufacturer that manufactures the product or the store owner who sells the product.
- the color determined by the color determination device 101 is the representative color of the product.
- the information on the representative colors of products can be used by online sales users when searching for products of a favorite color. That is, for example, when the user inputs a color and searches for products similar to the color.
- the color determination unit 105 determines a reference color closest to the representative color of the product among the predetermined reference colors as the classification color of the product.
- FIG. 8 is an explanatory diagram showing a schematic configuration of a product sales system according to this application example. Hereinafter, a description will be given with reference to FIG.
- the product sales system 501 shown in the figure is configured by connecting a terminal device 503, a product server 504, and a search server 505 via the Internet 502 so that they can communicate with each other.
- the product server 504 is a server device that realizes a so-called electronic market.
- the merchant who sells the merchandise accesses the merchandise server 504 via the browser operating on the terminal device 503 used by the merchant and registers the merchandise to be sold.
- a purchaser who purchases a product accesses the product server 504 via a browser operating on the terminal device 503 used by the purchaser, and searches for or purchases the product.
- the search server 505 corresponds to the image search device in the above embodiment.
- a document search function for searching a document in which related images are arranged is provided.
- the search server 505 may be realized by a server computer different from the product server 504, and the server computer that implements the product server 504 is configured to have the function of the search server 505. You can also. In the latter case, communication between the product server 504 and the search server 505 can be performed within the server computer without going through the Internet 502.
- the color determination device 101 and the color name determination device 301 are realized by the commodity server 504 according to the above embodiment executing a predetermined program.
- the administrator of the product server 504 selects a plurality of colors as reference colors in advance so that the purchaser can search for products by color.
- the database prepared in the product server 504 has a reference color table for managing the reference colors.
- the record registered in the reference color table has the following fields.
- Color ID This is an identification code uniquely assigned to each color in order to distinguish the reference color selected in advance by the administrator of the product server 504.
- Color name This is a character string that represents the color name of the reference color, and is given by the administrator of the product server 504.
- Color code The reference color is expressed by information corresponding to the coordinates of the color space. Typically, RGB representation is used.
- the number of colors prepared as reference colors is too small, it will not be useful for the purchaser's search, and if it is too large, it will be difficult for the purchaser to select a color during the search. Typically, the number is about 8 to 24 colors.
- the product server 504 manages products to be sold using a product table prepared in the database.
- the record registered in the product table has the following fields.
- Product ID In order to distinguish a product, it is an identification code that is uniquely assigned to each product, and is generally automatically generated each time registration is performed.
- Product name This is the name of the product.
- Brand name The name symbolizes a person who manufactures the product or a person who guarantees the quality of the product.
- Price This is the price when selling the product.
- Original color name This is a character string that represents the color of the product, and is given by the shop owner or manufacturer. It may be a compound word or a new word set independently. (6) Reference color ID.
- Store owner ID This is an identification code that is uniquely assigned to a merchant in order to identify the merchant who sells the product.
- Product image name An image file representing the product can be registered in the product server 504. The file name of the image file. When an image file is registered in a web server other than the product server 504, a URL (Universal Resource Locator) of the image file is used.
- URL Universal Resource Locator
- the product is obtained via the terminal device 503 by applying the technology of a normal electronic commerce system.
- the user who has accessed the server 504 can search for and purchase a product.
- FIG. 9A, 9B, and 9C are explanatory diagrams illustrating an example in which the product registration form transmitted from the product server 504 is displayed on the screen of the terminal device 503.
- FIG. 9C a description will be given with reference to FIG.
- a merchant ID column 702 As an input column for inputting information recorded in each field of the record of the product table, a merchant ID column 702, a product name column 703, a brand name column 704, a price A column 705 and a product image name column 706 are prepared.
- the store owner By observing the product image, the store owner inputs a character string representing the color of the product created by himself or a character string representing a color name provided by the manufacturer of the product into the original color name column 708. .
- a color name 711 of the reference color In the reference color ID column 709, a color name 711 of the reference color, a color sample 712 filled with the reference color, and a check box 713 for selecting the reference color are prepared.
- a check box 713 for selecting the reference color In FIG. 9A, since none of the reference colors is selected as a color for the product, none of the check boxes 713 is checked.
- the color of the pixel pointed to by the mouse cursor 751 is acquired. This color is transmitted to the product server 504.
- the merchandise server 504 functions as the color name determination device 301 that uses the search server 505 as a document search device, and returns a character string candidate representing the transmitted color to the terminal device 503.
- the terminal device 503 that has received the candidate for the character string representing the color displays a pop-up window 713 in the vicinity of the mouse cursor 751, as shown in FIG. 9C, and displays the received candidate therein.
- the shop owner can come up with an appropriate original color name with reference to this candidate.
- the candidate returned by the product server 504 may include the name of the reference color that is most similar to the color of the pixel designated by the mouse cursor 751.
- the product server 504 functions as the color determination device 101 that uses the search server 505 as an image search device, and represents a representative color represented by the transmitted original color name, a reference color similar to the representative color, To decide.
- the merchandise server 504 returns the color ID of the determined reference color to the terminal device 503.
- the terminal device 503 that has received the color ID includes a reference color name 711 corresponding to the received color ID and a color sample 712 filled with the reference color in the reference color ID column 709.
- the display mode of the check box 713 for selecting the reference color is changed, and the reference color is highlighted.
- the reference color is emphasized by enlarging the color name 711 and the color sample 712 displayed beside the check box 713.
- the store owner selects the most suitable one for the product with reference to the reference color being highlighted, and checks the check box 712 for the selected reference color.
- the merchant can make a correct classification by checking a check box 713 that is not highlighted.
- the reference color is wrongly determined, it may be considered that the selection of the original color name is almost inappropriate. In this case, a warning to that effect may be given to the shop owner, and the original color name may be re-selected.
- selectable colors may be limited to colors determined from the original color name.
- FIG. 9F there is also a mode in which a pop-up window 714 showing a color sample 712 of the reference color is displayed in the vicinity of the original color name column 708, and one of them is selected by the mouse. sell.
- each piece of information input to the product registration form 701 is transmitted to the product server 504 and registered as a record in the product table.
- the terminal device 503 itself can function as the color determination device 101 and the color name determination device 301 by appropriately configuring the script program. In this case, it is not necessary to transmit / receive information regarding the color as described above between the terminal device 503 and the product server 504.
- the first mode of use is to use a color ID when searching for products.
- FIG. 10 is an explanatory diagram showing an example in which the product search form transmitted from the product server 504 is displayed on the screen of the terminal device 503.
- a description will be given with reference to FIG.
- a color column 805 is prepared, and the user inputs these conditions.
- a search button 806 is clicked, a list of product descriptions 808 as search results is displayed in a search result field 807.
- the color column 805 is configured by a drop-down list box in which all the reference colors are arranged as selectable items, and the user can select a desired color from these.
- the second form of use is a case where a list of products of different colors of the product is displayed on a web page introducing a product.
- FIG. 11 is an explanatory diagram illustrating an example in which the product purchase form transmitted from the product server 504 is displayed on the screen of the terminal device 503.
- a description will be given with reference to FIG.
- the product purchase form 901 shown in this figure is displayed by clicking the product description 808 of the search result on the product search form 801.
- the product purchase form 901 includes a product name column, a brand name, a product description column 902 that displays a description by a store owner, a price column 903 that displays the price of the product, and a user purchase.
- a number field 904 for inputting a number and a purchase button 905 for starting a purchase process are arranged.
- the color sample 906 indicates that there are products of different colors displayed on the product purchase form 901.
- the different color product is a product whose merchant, product name, and brand name match but the reference color ID does not match.
- the browser shifts to a product purchase form for the different color product.
- the user can use colors in various ways.
- the color determination device 101 and the color name determination device 301 when the performance of the image search device is low, for example, an image search device that cannot provide an optimal solution for a query. Even if it is used, since the color appearance degree is acquired using a plurality of searched images, an appropriate color and color name can be proposed.
- a color name determination device it is possible to provide a color name determination device, a color name determination method, an information recording medium, and a program suitable for determining the correspondence between a color and a character string representing that color. .
Abstract
Description
色の入力を受け付ける色受付部、
前記受け付けられた色に関連する画像が配置された文書を検索した結果として得られた文書に関連するキーワード文字列を抽出するキーワード抽出部、
前記抽出されたキーワード文字列に関連する画像を検索した結果として得られた複数の画像のそれぞれに出現する色を所定の精度で表現し、当該所定の精度で表現された色が前記複数の画像のそれぞれに出現する度合を取得する度合取得部、
前記取得された度合に基づいて、前記所定の精度で表現された色から、前記受け付けられた文字列により表される色を決定する色決定部、
前記受け付けられた色と、前記決定された色と、が、同一または類似する場合、前記決定された色に対するキーワード文字列を、前記受け付けられた色を表す色名文字列に決定する色名決定部
を備えるように構成する。
前記検索の結果として得られる文書は複数であり、
前記キーワード抽出部は、前記検索された複数の文書に共通して出現するキーワード文字列を抽出する
ように構成することができる。
前記キーワード抽出部は、前記キーワード文字列を複数抽出し、
前記抽出されたキーワード文字列のそれぞれに対して、前記画像検索部による画像検索、前記度合取得部による度合の取得、前記色決定部による色の決定、前記色名決定部による色名の候補の決定が実行される
ように構成することができる。
前記文書検索では、前記受け付けられた色を入力したユーザにより指定された言語により記述された文書が検索される
ように構成することができる。
前記キーワード文字列は、前記指定された言語における名詞、名詞句、または、名詞を修飾する修飾語もしくは修飾語句である
ように構成することができる。
前記指定された言語において色を意味する文字列と、前記抽出されたキーワード文字列と、を指定するクエリを含む画像検索要求を、前記色名決定装置に通信可能に接続された外部の画像検索装置に送信し、当該画像検索装置から前記画像検索要求に対する画像検索応答を受信し、前記画像検索応答に指定される複数の画像を、画像検索の結果とする画像検索部
をさらに備えるように構成することができる。
前記抽出されたキーワード文字列に関連する画像を検索する画像検索部
をさらに備え、
前記度合取得部は、前記画像検索部による画像検索の結果として得られた複数の画像から、前記度合を取得する
ように構成することができる。
前記受け付けられた色を指定するクエリを含む文書検索要求を、前記色名決定装置に通信可能に接続された外部の文書検索装置に送信し、当該文書検索装置から前記文書検索要求に対する文書検索応答を受信し、前記文書検索応答に指定される文書を、文書検索の結果とする文書検索部
をさらに備えるように構成することができる。
前記受け付けられた色に関連する画像が配置された文書を検索する文書検索部
をさらに備え、
前記キーワード抽出部は、前記文書検索部による文書検索の結果として得られた文書に基づいて、前記キーワード文字列を抽出する
ように構成することができる。
色の入力を受け付ける色受付工程、
前記受け付けられた色に関連する画像が配置された文書を検索した結果として得られた文書に関連するキーワード文字列を抽出するキーワード抽出工程、
前記抽出されたキーワード文字列に関連する画像を検索した結果として得られた前記複数の画像のそれぞれに出現する色を所定の精度で表現し、当該所定の精度で表現された色が前記複数の画像のそれぞれに出現する度合を取得する度合取得工程、
前記取得された度合に基づいて、前記所定の精度で表現された色から、前記受け付けられた文字列により表される色を決定する色決定工程、
前記受け付けられた色と、前記決定された色と、が、同一または類似する場合、前記決定された色に対するキーワード文字列を、前記受け付けられた色を表す色名文字列に決定する色名決定工程
を備えるように構成する。
コンピュータを、
色の入力を受け付ける色受付部、
前記受け付けられた色に関連する画像が配置された文書を検索した結果として得られた文書に関連するキーワード文字列を抽出するキーワード抽出部、
前記抽出されたキーワード文字列に関連する画像を検索した結果として得られた複数の画像のそれぞれに出現する色を所定の精度で表現し、当該所定の精度で表現された色が前記複数の画像のそれぞれに出現する度合を取得する度合取得部、
前記取得された度合に基づいて、前記所定の精度で表現された色から、前記受け付けられた文字列により表される色を決定する色決定部、
前記受け付けられた色と、前記決定された色と、が、同一または類似する場合、前記決定された色に対するキーワード文字列を、前記受け付けられた色を表す色名文字列に決定する色名決定部
として機能させるプログラムを記録するように構成する。
コンピュータを、
色の入力を受け付ける色受付部、
前記受け付けられた色に関連する画像が配置された文書を検索した結果として得られた文書に関連するキーワード文字列を抽出するキーワード抽出部、
前記抽出されたキーワード文字列に関連する画像を検索した結果として得られた複数の画像のそれぞれに出現する色を所定の精度で表現し、当該所定の精度で表現された色が前記複数の画像のそれぞれに出現する度合を取得する度合取得部、
前記取得された度合に基づいて、前記所定の精度で表現された色から、前記受け付けられた文字列により表される色を決定する色決定部、
前記受け付けられた色と、前記決定された色と、が、同一または類似する場合、前記決定された色に対するキーワード文字列を、前記受け付けられた色を表す色名文字列に決定する色名決定部
として機能させるように構成する。
上記のように、色を定量的に表す体系として、RGB、CYMK、HSVなどの各種の色空間の定義があるが、これらの間での変換は定式化されている。以下では、理解を容易にするため、RGB表現を利用して説明するが、CYMKやHSVなど、他の表現を採用することも可能である。
従来から、インターネットなどのコンピュータ通信網で公開されている各種のウェブページや、個人用コンピュータ内、あるいは、社内LAN(Local Area Network)などで利用されている各種の文書を検索するための検索技術が提供され、各種のウェブ検索エンジンやデスクトップ検索エンジンが提供されている。
(1)文字列を与えると、その文字列により表現される色を決定する色決定装置、
(2)色を与えると、その色を表現する文字列(その色の名称、色名。)を決定する色名決定装置、
(3)上記(1)(2)を商品の色に応用した例
の順に、例をあげて説明する。
(1)検索結果の画像のうち、当該色が出現する画像の個数。
(2)検索結果の画像の画素のうち、当該色の画素の総数。
(3)上記(1)と(2)を組み合わせた値。すなわち、当該色が出現する画像の個数が多ければ多いほど、また、当該色の画素の総数が多ければ多いほど、大きくなるように定められる値。
(4)上記(1)(2)(3)のいずれかにおける値が所定の閾値以上の場合には、その値もしくは1を度合とし、所定の閾値未満の場合には、0を度合とする。
(1)度合が最も大きい色を、文字列により表される色とする。
(2)度合が最も大きい色が複数ある場合に、色空間において原点から最も遠い色を、文字列により表される色とする。RGB表現を採用している場合には、最も明るい色が選択されることになる。
(3)度合が最も大きい色が複数ある場合に、度合最大の色の平均色を、文字列により表される色とする。
(4)度合が大きい色の上位を所定個数選択し、その平均色を、文字列により表される色とする。
(5)各色について取得された度合を重みとして、色の重み付き平均をとり、その結果を文字列により表される色とする。
(1)度合が最も大きい色を、文字列により表される色とする。
(2)度合が大きい色の上位所定個数を、文字列により表される色とする。
のようなものを採用することができるが、これ以外の態様については、以降の実施例において説明する。
F[i,j,k] ← 1
を実行する。ここで、「←」は代入を意味する。また、フローチャートでは、「*」を用いることで範囲内のすべての添字に対して処理を行うことを表記する(以下同様)。
G[i,j,k] ← 0
を実行する。
i = (r×M)/255;
j = (g×M)/255;
k = (b×M)/255
を計算する。ここで、乗算「×」は除算「/」に優先して実行され、除算は整数除算であり、余りは切り捨てるものとする。
G[i,j,k] ← G[i,j,k] or 1
のように更新する(ステップS168)。ここで、「or」は、ビット論理和演算を意味する。
G[i,j,k] ← F[i,j,k] and G[i,j,k]
を実行する(ステップS170)。ここで「and」は、ビット論理積演算を意味する。
r = (i×255)/M;
g = (j×255)/M;
b = (k×255)/M
を計算すれば良い。
[i[0],j[0],k[0]],[i[1],j[1],k[1]],...,[i[L-1],j[L-1],k[L-1]]
r = (Σp=0 L-1 i[p]×255)/(M×L);
g = (Σp=0 L-1 j[p]×255)/(M×L);
b = (Σp=0 L-1 k[p]×255)/(M×L)
を計算すれば良い。
F[i,j,k] ← G[i,j,k]
を実行することと等価であるが、ビット配列全体のコピーや、変数F,Gが指すビット配列のアドレスを交換するなどの手法を採用することによって、処理の高速化を図ることもできる。
(1)RGB値やHSV値等、色空間における色成分の数値を、ユーザが直接入力する。
(2)ユーザが画像内の特定の画素を指示する。すると、当該画素の色が受け付けられる。任意の画像を利用することができるが、さまざまな色がグラデーションで描かれた色見本の画像などを採用しても良い。
(3)画像をユーザが指定するとその画像の代表色が受け付けられる。たとえば、指定された画像の背景の画素を実施例4の技術で除去した後、残った画素の平均色を代表色として利用する。
(1)検索結果の全文書において、当該候補が出現する頻度を求め、当該頻度が高いものをキーワード文字列として抽出する。
(2)検索結果の全文書のうち、当該候補が出現する文書の個数を求め、当該個数が高いものをキーワード文字列として抽出する。
(3)上記(1)(2)を組み合わせた評価値を求め、当該評価値が高いものをキーワード文字列として抽出する。たとえば、所定の係数による頻度と個数の線形和を、評価値として利用することができる。
(a)画素の色の分布と、
(b)分割数である正整数Kと、
が与えられると、
(x)K個の代表色と、
(y)クラスタの強さ(その代表色が表現するクラスタに属する画素の個数、分類された全画素数に対する当該クラスタに分類された画素数の割合等)と、
が出力される。日常的な慣習を考えると、ある色名が表現する色の個数は、3個程度までが適切と考えられる。このような場合には、Kとして3~6程度の値を指定してK-meansによりクラスタリングを行えば良い。
図8は、本応用例にかかる商品販売システムの概要構成を示す説明図である。以下、本図を参照して説明する。
商品サーバ504の管理者は、購入者が色で商品を検索できるようにするため、あらかじめ複数の色を基準色として選択する。商品サーバ504に用意されたデータベースには、この基準色を管理するための基準色テーブルが用意されている。
(1)カラーID。商品サーバ504の管理者があらかじめ選定した基準色を区別するために、色ごとに一意に割り振られる識別符号である。
(2)カラー名。当該基準色の色名を表す文字列であり、商品サーバ504の管理者が付与する。
(3)カラーコード。当該基準色を、色空間の座標に相当する情報で表現したもの。典型的には、RGB表現が利用される。
(1)商品ID。商品を区別するために、商品ごとに一意に割り振られる識別符号で、登録が行われるごとに、自動的に生成されるのが一般的である。
(2)商品名。当該商品の呼称である。
(3)ブランド名。当該商品を製造している者や、当該商品の品質を保証している者などを象徴する名称である。
(4)価格。当該商品を販売する際の価格である。
(5)オリジナルカラー名。当該商品の色を表現する文字列であり、商店主や製造者が付与する。合成語や、独自に設定された新語などのこともある。
(6)基準カラーID。あらかじめ定めた基準色のいずれに、当該商品の色が類似するかを表す。
(7)商店主ID。当該商品を販売する商店主を識別するために、一意に商店主に割り振られる識別符号である。
(8)商品画像名。商品サーバ504には、当該商品を表す画像ファイルを登録することができる。その画像ファイルのファイル名である。なお、商品サーバ504以外のウェブサーバに画像ファイルを登録した場合には、当該画像ファイルのURL(Universal Resource Locator)を使用する。
(9)その他の商品に関する関連情報。電子商取引に係る種々の情報が入れられる。
図9A、図9B、図9Cは、商品サーバ504から送信された商品登録フォームが、端末装置503の画面に表示される例を示す説明図である。以下、本図を参照して説明する。
本応用例では、商品テーブルの各レコードに、当該レコードの商品に最も適合した基準色のカラーIDの情報が登録されることになる。以下では、この情報の利用の態様について説明する。
102 文字列受付部
103 画像検索部
104 度合取得部
105 色決定部
301 色名決定装置
302 色受付部
303 文書検索部
304 キーワード抽出部
305 色名決定部
501 商品販売システム
502 インターネット
503 端末装置
504 商品サーバ
505 検索サーバ
701 商品登録フォーム
702 商店主ID欄
703 商品名欄
704 ブランド名欄
705 価格欄
706 商品画像名欄
707 商品画像欄
708 オリジナルカラー名欄
709 基準カラーID欄
711 カラー名
712 カラー見本
713 チェックボックス
714 ポップアップウィンドウ
721 登録ボタン
751 マウスカーソル
801 商品検索フォーム
802 商品名キーワード欄
803 ブランド名キーワード欄
804 価格帯欄
805 カラー欄
806 検索ボタン
807 検索結果欄
808 商品説明
901 商品購入フォーム
902 商品説明欄
903 価格欄
904 個数欄
905 購入ボタン
906 カラー見本
Claims (12)
- 色の入力を受け付ける色受付部、
前記受け付けられた色に関連する画像が配置された文書を検索した結果として得られた文書に関連するキーワード文字列を抽出するキーワード抽出部、
前記抽出されたキーワード文字列に関連する画像を検索した結果として得られた複数の画像のそれぞれに出現する色を所定の精度で表現し、当該所定の精度で表現された色が前記複数の画像のそれぞれに出現する度合を取得する度合取得部、
前記取得された度合に基づいて、前記所定の精度で表現された色から、前記受け付けられた文字列により表される色を決定する色決定部、
前記受け付けられた色と、前記決定された色と、が、同一または類似する場合、前記決定された色に対するキーワード文字列を、前記受け付けられた色を表す色名文字列に決定する色名決定部
を備えることを特徴とする色名決定装置。 - 請求項1に記載の色名決定装置であって、
前記検索の結果として得られる文書は複数であり、
前記キーワード抽出部は、前記検索された複数の文書に共通して出現するキーワード文字列を抽出する
ことを特徴とする色名決定装置。 - 請求項1に記載の色名決定装置であって、
前記キーワード抽出部は、前記キーワード文字列を複数抽出し、
前記抽出されたキーワード文字列のそれぞれに対して、前記画像検索部による画像検索、前記度合取得部による度合の取得、前記色決定部による色の決定、前記色名決定部による色名の候補の決定が実行される
ことを特徴とする色名決定装置。 - 請求項1に記載の色名決定装置であって、
前記文書検索では、前記受け付けられた色を入力したユーザにより指定された言語により記述された文書が検索される
ことを特徴とする色名決定装置。 - 請求項4に記載の色名決定装置であって、
前記キーワード文字列は、前記指定された言語における名詞、名詞句、または、名詞を修飾する修飾語もしくは修飾語句である
ことを特徴とする色名決定装置。 - 請求項4に記載の色名決定装置であって、
前記指定された言語において色を意味する文字列と、前記抽出されたキーワード文字列と、を指定するクエリを含む画像検索要求を、前記色名決定装置に通信可能に接続された外部の画像検索装置に送信し、当該画像検索装置から前記画像検索要求に対する画像検索応答を受信し、前記画像検索応答に指定される複数の画像を、画像検索の結果とする画像検索部
をさらに備えることを特徴とする色名決定装置。 - 請求項1に記載の色名決定装置であって、
前記抽出されたキーワード文字列に関連する画像を検索する画像検索部
をさらに備え、
前記度合取得部は、前記画像検索部による画像検索の結果として得られた複数の画像から、前記度合を取得する
ことを特徴とする色名決定装置。 - 請求項1に記載の色名決定装置であって、
前記受け付けられた色を指定するクエリを含む文書検索要求を、前記色名決定装置に通信可能に接続された外部の文書検索装置に送信し、当該文書検索装置から前記文書検索要求に対する文書検索応答を受信し、前記文書検索応答に指定される文書を、文書検索の結果とする文書検索部
をさらに備えることを特徴とする色名決定装置。 - 請求項1に記載の色名決定装置であって、
前記受け付けられた色に関連する画像が配置された文書を検索する文書検索部
をさらに備え、
前記キーワード抽出部は、前記文書検索部による文書検索の結果として得られた文書に基づいて、前記キーワード文字列を抽出する
ことを特徴とする色名決定装置。 - 色の入力を受け付ける色受付工程、
前記受け付けられた色に関連する画像が配置された文書を検索した結果として得られた文書に関連するキーワード文字列を抽出するキーワード抽出工程、
前記抽出されたキーワード文字列に関連する画像を検索した結果として得られた前記複数の画像のそれぞれに出現する色を所定の精度で表現し、当該所定の精度で表現された色が前記複数の画像のそれぞれに出現する度合を取得する度合取得工程、
前記取得された度合に基づいて、前記所定の精度で表現された色から、前記受け付けられた文字列により表される色を決定する色決定工程、
前記受け付けられた色と、前記決定された色と、が、同一または類似する場合、前記決定された色に対するキーワード文字列を、前記受け付けられた色を表す色名文字列に決定する色名決定工程
を備えることを特徴とする色名決定方法。 - コンピュータを、
色の入力を受け付ける色受付部、
前記受け付けられた色に関連する画像が配置された文書を検索した結果として得られた文書に関連するキーワード文字列を抽出するキーワード抽出部、
前記抽出されたキーワード文字列に関連する画像を検索した結果として得られた複数の画像のそれぞれに出現する色を所定の精度で表現し、当該所定の精度で表現された色が前記複数の画像のそれぞれに出現する度合を取得する度合取得部、
前記取得された度合に基づいて、前記所定の精度で表現された色から、前記受け付けられた文字列により表される色を決定する色決定部、
前記受け付けられた色と、前記決定された色と、が、同一または類似する場合、前記決定された色に対するキーワード文字列を、前記受け付けられた色を表す色名文字列に決定する色名決定部
として機能させることを特徴とするプログラムを記録したコンピュータ読み取り可能な情報記録媒体。 - コンピュータを、
色の入力を受け付ける色受付部、
前記受け付けられた色に関連する画像が配置された文書を検索した結果として得られた文書に関連するキーワード文字列を抽出するキーワード抽出部、
前記抽出されたキーワード文字列に関連する画像を検索した結果として得られた複数の画像のそれぞれに出現する色を所定の精度で表現し、当該所定の精度で表現された色が前記複数の画像のそれぞれに出現する度合を取得する度合取得部、
前記取得された度合に基づいて、前記所定の精度で表現された色から、前記受け付けられた文字列により表される色を決定する色決定部、
前記受け付けられた色と、前記決定された色と、が、同一または類似する場合、前記決定された色に対するキーワード文字列を、前記受け付けられた色を表す色名文字列に決定する色名決定部
として機能させることを特徴とするプログラム。
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US14/235,523 US8891860B2 (en) | 2011-08-05 | 2012-08-01 | Color name determination device, color name determination method, information recording medium, and program |
JP2013511432A JP5259894B1 (ja) | 2011-08-05 | 2012-08-01 | 色名決定装置、色名決定方法、情報記録媒体、ならびに、プログラム |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2011-172211 | 2011-08-05 | ||
JP2011172211 | 2011-08-05 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2013021889A1 true WO2013021889A1 (ja) | 2013-02-14 |
Family
ID=47668399
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2012/069565 WO2013021889A1 (ja) | 2011-08-05 | 2012-08-01 | 色名決定装置、色名決定方法、情報記録媒体、ならびに、プログラム |
Country Status (3)
Country | Link |
---|---|
US (1) | US8891860B2 (ja) |
JP (1) | JP5259894B1 (ja) |
WO (1) | WO2013021889A1 (ja) |
Families Citing this family (26)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9460157B2 (en) * | 2012-12-28 | 2016-10-04 | Wal-Mart Stores, Inc. | Ranking search results based on color |
US9460214B2 (en) | 2012-12-28 | 2016-10-04 | Wal-Mart Stores, Inc. | Ranking search results based on color |
WO2015145766A1 (ja) * | 2014-03-28 | 2015-10-01 | 楽天株式会社 | 色推定装置、色推定方法及び色推定プログラム |
US9727983B2 (en) | 2014-06-26 | 2017-08-08 | Amazon Technologies, Inc. | Automatic color palette based recommendations |
US10255295B2 (en) | 2014-06-26 | 2019-04-09 | Amazon Technologies, Inc. | Automatic color validation of image metadata |
US10235389B2 (en) | 2014-06-26 | 2019-03-19 | Amazon Technologies, Inc. | Identifying data from keyword searches of color palettes |
US10223427B1 (en) | 2014-06-26 | 2019-03-05 | Amazon Technologies, Inc. | Building a palette of colors based on human color preferences |
US9922050B2 (en) | 2014-06-26 | 2018-03-20 | Amazon Technologies, Inc. | Identifying data from keyword searches of color palettes and color palette trends |
US9916613B1 (en) | 2014-06-26 | 2018-03-13 | Amazon Technologies, Inc. | Automatic color palette based recommendations for affiliated colors |
US9401032B1 (en) | 2014-06-26 | 2016-07-26 | Amazon Technologies, Inc. | Image-based color palette generation |
US9659032B1 (en) | 2014-06-26 | 2017-05-23 | Amazon Technologies, Inc. | Building a palette of colors from a plurality of colors based on human color preferences |
US9697573B1 (en) | 2014-06-26 | 2017-07-04 | Amazon Technologies, Inc. | Color-related social networking recommendations using affiliated colors |
US9996579B2 (en) | 2014-06-26 | 2018-06-12 | Amazon Technologies, Inc. | Fast color searching |
US10691744B2 (en) | 2014-06-26 | 2020-06-23 | Amazon Technologies, Inc. | Determining affiliated colors from keyword searches of color palettes |
US9679532B2 (en) | 2014-06-26 | 2017-06-13 | Amazon Technologies, Inc. | Automatic image-based recommendations using a color palette |
US9514543B2 (en) | 2014-06-26 | 2016-12-06 | Amazon Technologies, Inc. | Color name generation from images and color palettes |
US10073860B2 (en) | 2014-06-26 | 2018-09-11 | Amazon Technologies, Inc. | Generating visualizations from keyword searches of color palettes |
US9792303B2 (en) | 2014-06-26 | 2017-10-17 | Amazon Technologies, Inc. | Identifying data from keyword searches of color palettes and keyword trends |
US9898487B2 (en) | 2014-06-26 | 2018-02-20 | Amazon Technologies, Inc. | Determining color names from keyword searches of color palettes |
US10169803B2 (en) | 2014-06-26 | 2019-01-01 | Amazon Technologies, Inc. | Color based social networking recommendations |
US9652868B2 (en) | 2014-06-26 | 2017-05-16 | Amazon Technologies, Inc. | Automatic color palette based recommendations |
US10120880B2 (en) * | 2014-06-26 | 2018-11-06 | Amazon Technologies, Inc. | Automatic image-based recommendations using a color palette |
US9524563B2 (en) * | 2014-06-26 | 2016-12-20 | Amazon Technologies, Inc. | Automatic image-based recommendations using a color palette |
US10430857B1 (en) | 2014-08-01 | 2019-10-01 | Amazon Technologies, Inc. | Color name based search |
US9785649B1 (en) | 2014-09-02 | 2017-10-10 | Amazon Technologies, Inc. | Hue-based color naming for an image |
US11436771B2 (en) | 2020-11-20 | 2022-09-06 | International Business Machines Corporation | Graph-based color description generation |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH07146871A (ja) * | 1993-11-24 | 1995-06-06 | Hitachi Ltd | 静止画検索装置および静止画検索方法 |
JP2000112959A (ja) * | 1998-09-30 | 2000-04-21 | Canon Inc | 情報検索装置及び方法 |
JP2005202452A (ja) * | 2004-01-13 | 2005-07-28 | Casio Comput Co Ltd | 画像表示装置、画像記録装置及びそのプログラム |
JP2009032154A (ja) * | 2007-07-30 | 2009-02-12 | Olympus Corp | 情報処理装置およびプログラム |
Family Cites Families (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3748164B2 (ja) * | 1998-02-06 | 2006-02-22 | 富士通株式会社 | パターン抽出装置 |
JP2001069362A (ja) | 1999-08-27 | 2001-03-16 | Minolta Co Ltd | 色補正装置 |
KR100422805B1 (ko) * | 2000-12-26 | 2004-03-12 | 한국전자통신연구원 | 자연어를 이용한 이미지 데이터 색상 및 형태 검색 장치및 그 방법 |
US7693856B2 (en) * | 2004-06-25 | 2010-04-06 | Apple Inc. | Methods and systems for managing data |
JP4960897B2 (ja) * | 2008-01-30 | 2012-06-27 | 株式会社リコー | 画像処理装置、画像処理方法、プログラム、記憶媒体 |
JP5682569B2 (ja) * | 2009-10-16 | 2015-03-11 | 日本電気株式会社 | 色彩解析装置、色彩解析方法、及び色彩解析プログラム |
JP5259876B2 (ja) * | 2010-02-26 | 2013-08-07 | 楽天株式会社 | 情報処理装置、情報処理方法、情報処理プログラムを記録した記録媒体 |
US8861844B2 (en) * | 2010-03-29 | 2014-10-14 | Ebay Inc. | Pre-computing digests for image similarity searching of image-based listings in a network-based publication system |
-
2012
- 2012-08-01 WO PCT/JP2012/069565 patent/WO2013021889A1/ja active Application Filing
- 2012-08-01 JP JP2013511432A patent/JP5259894B1/ja active Active
- 2012-08-01 US US14/235,523 patent/US8891860B2/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH07146871A (ja) * | 1993-11-24 | 1995-06-06 | Hitachi Ltd | 静止画検索装置および静止画検索方法 |
JP2000112959A (ja) * | 1998-09-30 | 2000-04-21 | Canon Inc | 情報検索装置及び方法 |
JP2005202452A (ja) * | 2004-01-13 | 2005-07-28 | Casio Comput Co Ltd | 画像表示装置、画像記録装置及びそのプログラム |
JP2009032154A (ja) * | 2007-07-30 | 2009-02-12 | Olympus Corp | 情報処理装置およびプログラム |
Also Published As
Publication number | Publication date |
---|---|
JPWO2013021889A1 (ja) | 2015-03-05 |
US8891860B2 (en) | 2014-11-18 |
JP5259894B1 (ja) | 2013-08-07 |
US20140177952A1 (en) | 2014-06-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP5259893B1 (ja) | 色決定装置、色決定システム、色決定方法、情報記録媒体、ならびに、プログラム | |
JP5259894B1 (ja) | 色名決定装置、色名決定方法、情報記録媒体、ならびに、プログラム | |
Wang et al. | Recipe recognition with large multimodal food dataset | |
US10691744B2 (en) | Determining affiliated colors from keyword searches of color palettes | |
US8234306B2 (en) | Information process apparatus, information process method, and program | |
US20110078176A1 (en) | Image search apparatus and method | |
Wang et al. | Affective image adjustment with a single word | |
CN107145536A (zh) | 用户画像构建方法与装置及推荐方法与装置 | |
US11615263B2 (en) | Content prediction based on pixel-based vectors | |
JP6767342B2 (ja) | 検索装置、検索方法および検索プログラム | |
US11403339B2 (en) | Techniques for identifying color profiles for textual queries | |
JP6310529B1 (ja) | 検索装置、検索方法および検索プログラム | |
Maheshwari et al. | Learning colour representations of search queries | |
JP2000048041A (ja) | データ検索システム及びこれに用いる装置 | |
JP6756648B2 (ja) | 生成装置、生成方法および生成プログラム | |
JP5153390B2 (ja) | 関連語辞書作成方法及び装置、並びに関連語辞書作成プログラム | |
JP7372278B2 (ja) | 算出装置、算出方法及び算出プログラム | |
Lin et al. | Robustness and reliability evaluations of image annotation | |
US20220019617A1 (en) | Image processing device, image viewing system, image processing program, and image processing method | |
WO2023062668A1 (ja) | 情報処理装置、情報処理方法、情報処理システム、およびプログラム | |
JP2012108721A (ja) | 画像検索装置、及び画像検索方法 | |
JP7027288B2 (ja) | 決定装置、決定方法および決定プログラム | |
JP6805626B2 (ja) | 情報処理装置及びプログラム | |
Xu | Cross-Media Retrieval: Methodologies and Challenges | |
JP2022178243A (ja) | 画像生成装置、画像生成方法およびプログラム |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
ENP | Entry into the national phase |
Ref document number: 2013511432 Country of ref document: JP Kind code of ref document: A |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 12822782 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 14235523 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 12822782 Country of ref document: EP Kind code of ref document: A1 |