CN103699532A - Image color retrieval method and system - Google Patents
Image color retrieval method and system Download PDFInfo
- Publication number
- CN103699532A CN103699532A CN201210365623.XA CN201210365623A CN103699532A CN 103699532 A CN103699532 A CN 103699532A CN 201210365623 A CN201210365623 A CN 201210365623A CN 103699532 A CN103699532 A CN 103699532A
- Authority
- CN
- China
- Prior art keywords
- value
- pixel
- color
- tone
- quantized
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/5838—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using colour
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Library & Information Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses an image color retrieval method and an image color retrieval system and relates to the field of image processing. The method comprises the steps of performing quantification processing to HSV (Hue, Saturation and Value) color spaces of pixels in an image to obtain quantified color values within a range of 0 to N, wherein non-uniform fuzzy quantification is performed to tone values; statistically collecting a histogram of the quantified color values of the pixel of the image and performing normalization processing by using the sum of the pixels of the image to obtain a color characteristic vector of the image; comparing the color characteristic vector of the image to the color characteristic vectors of other images to determine the similarity of the images. The method and the system disclosed by the invention have the advantages that since images related to the inquired image color are extracted from a database by taking the color characteristic vector after image quantification as a relevant judgment basis, the image retrieval is enabled to be more intelligent and the extraction result of color-related images becomes more accurate; since the images in the database are processed in advance by adopting an offline module, the execution speed is effectively improved.
Description
Technical field
The present invention relates to technical field of image processing, particularly a kind of color of image search method and system.
Background technology
Fast development along with multimedia technology, digital imaging technology and computer technology application, and mass-memory unit and digitizer popularization and application, image and video data present the geometric trend that increases progressively, and meanwhile, have occurred jumbo image/video data storehouse.But digital imaging apparatus itself is not to image database management and query function, how from immense image data base, finding rapidly, accurately needed digital picture has become one of focus of multimedia technology research in recent years.The research of image retrieval technologies provides strong support to a plurality of research fields such as multimedia digital library, trade mark copyright management and satellite remote sensing Geographic Information System.
As far back as phase late 1970s, people have just proposed the method for image retrieval, and this search method is a kind of retrieval technique based on text/data structure, by the keyword of Description Image or free text are retrieved.First need the view data in database to carry out textual description, the process of inquiry is accurate coupling or the probability match based on iamge description text.Yet, this image search method has the following shortcoming that is difficult to overcome: 1) very strong by the means subjectivity manually marking, the different people of same sub-picture may have different explanations to its content, same description vocabulary also may have various meaning, and country variant different nationalities is difficult to a kind of unified language, image be made an explanation.2) expansion due to data scale causes the expense of artificial mark to be more and more difficult to bear.3) content of image is abundanter, has the feature of " a figure K word ", in some cases, with limited word, cannot clearly express the expressed meaning of image itself at all, especially when image implication has certain connotation.4) although a lot of parameter informations such as the date created of image, format information, resolution, can by robotization, be extracted to a certain extent, and some basic index clues are provided, but the method marking based on content of text is difficult to realize on to the similarity retrieval of Image Visual Feature.
Summary of the invention
The present inventor finds to have problems in above-mentioned prior art, and therefore at least one problem in described problem, has proposed a kind of new technical scheme.
An object of the present invention is to provide a kind of for for carrying out quickly and accurately the technical scheme of image retrieval.
According to a first aspect of the invention, provide a kind of image search method, having comprised: the tone, saturation degree and the brightness hsv color space representation that obtain pixel in image; Hsv color space representation to pixel in described image carries out quantification treatment, obtains and belongs to the quantized color value within the scope of 0 ~ N, and N is natural number, wherein, tone value is carried out to non-homogeneous fuzzy quantization; The histogram of adding up the quantized color value of described image pixel, is normalized with the sum of all pixels of described image, obtains the color feature vector of described image; The color feature vector of the color feature vector of described image and other images is compared to determine to the similarity of image.
Alternatively, tone value is carried out to non-homogeneous fuzzy quantization to be comprised: the intersection at tone arranges zone of transition, if tone value drops in described zone of transition, described tone value is quantified as to the weighted sum of the color quantization value relevant to described zone of transition, if described tone value does not drop in described zone of transition, described tone value is directly quantized.
Alternatively, tone value being carried out to non-homogeneous fuzzy quantization comprises: a plurality of quantization boundaries that tone is set; On the both sides of described quantization boundary, quantization boundary zone of transition is set, the size of described zone of transition and the range size between two adjacent quantization boundaries are directly proportional; Obtain the membership function of described zone of transition, according to ridge type, distribute, try to achieve the membership function value that tone value belongs to zone of transition; If the tone value of pixel drops in described zone of transition, the tone value of described pixel is quantified as to the weighted sum of two the color quantization values relevant to this zone of transition, weighted value is membership function value; If the tone value of pixel drops on outside zone of transition, according to tone value, determine corresponding quantized value.
Alternatively, N is 36, and described quantization boundary comprises 22,45,70,155,186,278,330.
Alternatively, the method also comprises: the pixel of described image is represented to be converted to hsv color space representation by RGB color space, and wherein, the span of tone is 0 to 360, and the span of saturation degree is 0 to 1, and the span of brightness is 0 to 1.
Alternatively, the hsv color space representation of pixel in described image being carried out to quantification treatment obtains the quantized color value belong within the scope of 0 ~ N and comprises: if the brightness value of pixel is less than 0.2, the quantized color value value of pixel is 0; If the brightness value of pixel is more than or equal to 0.8, the value of saturation degree is less than 0.2, and the quantized color value value of pixel is 7; If the brightness value of pixel is between 0.2 and 0.8, the value of saturation degree s is less than 0.2, the value after 10 demultiplications 1 that the quantized color value value of pixel is brightness value also round downwards; If the brightness value of pixel is more than or equal to 0.2, the value of saturation degree is more than or equal to 0.2, and the quantized color value of pixel is obtained by the quantized value weighting of brightness value, intensity value and the tone value of pixel.
Alternatively, if the brightness value of pixel is more than or equal to 0.2, the value of saturation degree is more than or equal to 0.2, the quantized color value of pixel is obtained and comprised by the quantized value weighting of brightness value, intensity value and the tone value of pixel: if the brightness value of pixel between 0.2 and 0.7, the quantized value of the brightness value of pixel is 0; If the value of brightness v is between 0.7 and 1, the quantized value of the brightness value of pixel is 1; If the intensity value of pixel is between 0.2 and 0.65, the quantized value of the intensity value of pixel is 0; If the intensity value of pixel is between 0.65 and 1, the quantized value of the intensity value of pixel is 1; The tone value of pixel is carried out to non-homogeneous fuzzy quantization processing; The quantized color value of pixel is obtained by the quantized value weighting of tone value, intensity value and the brightness value of pixel: I=4H+2S+V+8, and wherein, I is quantized color value, and H is tone, and S is saturation degree, and V is brightness.
Alternatively, the color feature vector of the color feature vector of described image and other images is compared to determine that similar image comprises: the color feature vector of the color feature vector of described image and other images is compared, according to the absolute value distance between color feature vector or Euclidean distance, calculate the similarity of described image and other images.
According to a further aspect in the invention, provide a kind of image indexing system, comprising: color space represents acquisition module, for obtaining tone, saturation degree and the brightness hsv color space representation of image pixel; Color space represents quantization modules, for the hsv color space representation to described image pixel, carries out quantification treatment, obtains and belongs to the quantized color value within the scope of 0 ~ N, and N is natural number, wherein, tone value is carried out to non-homogeneous fuzzy quantization; Color feature vector acquisition module, for adding up the histogram of the quantized color value of described image pixel, is normalized with the sum of all pixels of described image, obtains the color feature vector of described image; Color feature vector comparison module, for comparing the color feature vector of the color feature vector of described image and other images to determine the similarity of image.
Alternatively, described color space represents that quantization modules comprises: luminance quantization unit, for brightness value being quantized to obtain luminance quantization value; Saturation degree quantifying unit, for quantizing to obtain saturation degree quantized value to intensity value; Tone quantifying unit, obtains tone quantized value for tone value being carried out to non-homogeneous fuzzy quantization; Quantized color value determining unit, for determining the quantized color value that belongs to the pixel within the scope of 0 ~ N according to the brightness value of pixel, intensity value and described luminance quantization value, saturation degree quantized value and tone quantized value.
Alternatively, tone quantifying unit arranges zone of transition at the intersection of tone, if tone value drops in described zone of transition, described tone value is quantified as to the weighted sum of the color quantization value relevant to described zone of transition, if described tone value does not drop in described zone of transition, described tone value is directly quantized.
Alternatively, tone quantifying unit is for arranging a plurality of quantization boundaries of tone; On the both sides of described quantization boundary, quantization boundary zone of transition is set, the size of described zone of transition and the range size between two adjacent quantization boundaries are directly proportional; Obtain the membership function of described zone of transition, according to ridge type, distribute, try to achieve the membership function value that tone value belongs to zone of transition; If the tone value of pixel drops in described zone of transition, the tone value of described pixel is quantified as to the weighted sum of two the color quantization values relevant to this zone of transition, weighted value is membership function value; If the tone value of pixel drops on outside zone of transition, directly according to tone value, determine corresponding quantized value.
Alternatively, N is 36, and described quantization boundary comprises 22,45,70,155,186,278,330.
Alternatively, if the brightness value of quantized color value determining unit judgement pixel is less than 0.2, the quantized color value value of pixel is 0; If the brightness value of pixel is greater than 0.8, the value of saturation degree is less than 0.2, and the quantized color value value of pixel is 7; If the brightness value of pixel is between 0.2 and 0.8, the value of saturation degree is less than 0.2, the value after 10 demultiplications 1 that the quantized color value value of pixel is brightness value also round downwards; If the brightness value of pixel is more than or equal to 0.2, the value of saturation degree is more than or equal to 0.2, and the weighting of the tone quantized value that the quantized color value of pixel is obtained by the luminance quantization value of described luminance quantization unit acquisition pixel, described saturation degree quantifying unit acquisition saturation degree quantized value and described tone quantifying unit obtains the quantized color value of pixel.
Alternatively, if the brightness value of pixel is more than or equal to 0.2, the value of saturation degree s is more than or equal to 0.2, if the brightness value of described luminance quantization unit judges pixel is between 0.2 and 0.7, the quantized value of the brightness value of pixel is 0, if the value of brightness v is between 0.7 and 1, the quantized value of the brightness value of pixel is 1; If the intensity value of described saturation degree quantifying unit judgement pixel is between 0.2 and 0.65, the quantized value of the intensity value of pixel is 0, if the intensity value of pixel between 0.65 and 1, the quantized value of the intensity value of pixel is 1; Described tone quantifying unit is carried out non-homogeneous fuzzy quantization processing to the tone value of pixel.
Alternatively, color space represents acquisition module, and the pixel of described image is represented to be converted to hsv color space representation by RGB color space, and wherein, the span of tone is 0 to 360, and the span of saturation degree is 0 to 1, and the span of brightness is 0 to 1.
Alternatively, color feature vector comparison module compares the color feature vector of the color feature vector of described image and other images, calculates the similarity of described image and other images according to the absolute value distance between color feature vector or Euclidean distance.。
An advantage of the present invention is, the color feature vector of usining after image quantization, as the criterion of image similarity, has directly utilized the content information of image, without manually image being described, make image retrieval quicker, the extraction result of color associated picture is also more accurate.
By the detailed description to exemplary embodiment of the present invention referring to accompanying drawing, it is clear that further feature of the present invention and advantage thereof will become.
Accompanying drawing explanation
The accompanying drawing that forms a part for instructions has been described embodiments of the invention, and together with the description for explaining principle of the present invention.
With reference to accompanying drawing, according to detailed description below, can more be expressly understood the present invention, wherein:
Fig. 1 illustrates the process flow diagram of an embodiment of image search method of the present invention.
Fig. 2 illustrates general structure and the workflow diagram of an embodiment of image indexing system of the present invention.
Fig. 3 illustrates the process flow diagram of an example of color space conversion of the present invention.
Fig. 4 illustrates color quantizing main flow chart in one embodiment of the present of invention.
Fig. 5 illustrates colored area quantization process flow diagram in one embodiment of the present of invention.
Fig. 6 illustrates tone fuzzy quantization process flow diagram in one embodiment of the present of invention.
Fig. 7 illustrates the structural drawing of an embodiment of image indexing system of the present invention.
Fig. 8 illustrates the structural drawing of another embodiment of image indexing system of the present invention.
Embodiment
Now with reference to accompanying drawing, describe various exemplary embodiment of the present invention in detail.It should be noted that: unless illustrate in addition, the parts of setting forth in these embodiments and positioned opposite, numeral expression formula and the numerical value of step do not limit the scope of the invention.
, it should be understood that for convenience of description, the size of the various piece shown in accompanying drawing is not to draw according to actual proportionate relationship meanwhile.
To the description only actually of at least one exemplary embodiment, be illustrative below, never as any restriction to the present invention and application or use.
For the known technology of person of ordinary skill in the relevant, method and apparatus, may not discuss in detail, but in suitable situation, described technology, method and apparatus should be regarded as authorizing a part for instructions.
In all examples with discussing shown here, it is exemplary that any occurrence should be construed as merely, rather than as restriction.Therefore, other example of exemplary embodiment can have different values.
It should be noted that: in similar label and letter accompanying drawing below, represent similar terms, therefore, once be defined in an a certain Xiang Yi accompanying drawing, in accompanying drawing subsequently, do not need it to be further discussed.
Fig. 1 illustrates the process flow diagram of an embodiment of image search method of the present invention.
As shown in Figure 1, step 102, obtains tone (Hue) h, saturation degree (Saturation) s of pixel and the hsv color space representation of brightness (Value) v in image.The color space of image has a variety of, and for example RGB, YIQ, YUV, YCbCr, YES also have color space HIS, HSV by color psychology three attribute representations towards tone, HSL, HSB, TSL etc.Hsv color space representation adopts tone H, the saturation degree S mode separated with brightness V to realize the quantitative description to color, has reflected more exactly the understanding mode of human visual system to color, better to image area calibration.For the image representing by other color spaces, can be converted to HSV by the conversion formula between each color space.Concrete conversion realizes can be referring to the introduction of correlation technique document, for not being described in detail at this for purpose of brevity.
In above-described embodiment, the color feature vector of usining after image quantization, as relevant criterion, extracts the image relevant to query image color in database.Directly utilized the content information of image, without manually image being described, made image indexing system more intelligent, the extraction result of color associated picture is also more accurate.Adopt the mode of fuzzy quantization to carry out quantification treatment to tone h, fully take into account similarity and the continuity of tone h quantization boundary vicinity color, reduce quantization error.
Fig. 2 illustrates the workflow diagram of an embodiment of image indexing system of the present invention.In this embodiment, this system comprises processed offline module 21 and two modules 22 of online processing.
Wherein, processed offline module 21 is responsible for the image in database to process, for the every piece image in database produces corresponding color feature vector, by the color feature vector sum image of image together corresponding stored in database.While carrying out follow-up feature comparison, can directly import the color feature vector of image, without image being carried out to the re-treatment of proper vector extraction, to improve the execution speed of system.The workflow of processed offline module 21 comprises that database images reads 211a, color of image space conversion 212a, color fuzzy quantization is processed 213a, color feature 214a and generates color feature vector file.Be specifically described as follows:
Step 211a, database images reads.Obtain the title of image in database, form image name list.In can computational data storehouse comprise image sum M, and use 1～M image in registration database respectively, also can represent each image according to other unique identifications of image.According to mark mark, successively the image in database is read in to processed offline module.The image reading in is for example used RGB color space to represent in system.
Step 212a, the conversion of color of image space.Compare RGB color space, HSV is digitized processing color better.Before quantizing, first image is carried out to color space conversion, by rgb space, be transformed into HSV space.After color of image space converts, each pixel of image is used tone h, saturation degree s and brightness v to represent, wherein the span of h is that 0 to 360, s span is that 0 to 1, v span is 0 to 1.
Step 213a, color fuzzy quantization is processed.The color of coloured image is enriched very much, and a width rgb image for example can have the color that 256*256*256 kind is different, and while directly carrying out color comparison, calculated amount is very large.Consider that human eye is limited to the resolution of color, within the scope of certain, human eye even cannot be perceiveed the variation of color is insensitive, can use the mode of color quantizing to reduce color sum, thereby reduce the dimension of color characteristic, shorten the computing time of color comparison.Generally speaking, in visual signature, the difference between color is mainly embodied by the difference of shade of color.For reducing computation complexity, when carrying out color quantizing, only tone h component is carried out to fuzzy quantization, s and v component adopt the mode directly quantizing to process.So-called fuzzy quantization is exactly that intersection at tone arranges zone of transition, for given h value, if dropped in zone of transition, think that the color quantizing of this pixel exists ambiguity, just it is quantified as to the weighted sum of the several color quantization values relevant to this zone of transition, for the pixel not dropping in zone of transition, adopts the method for direct quantification to quantize.After image is carried out to the processing of color fuzzy quantization, after each pixel quantizes, the span of color value is between 0 ~ N-1.
Step 214a, color feature.Color value in statistic quantification image is respectively 0 to N-1 pixel quantity, and is normalized with the sum of all pixels of image, has just formed the N dimension color feature vector of this image.
Image in database, after above-mentioned steps is finished dealing with, is saved in database images color feature vector file together with the title of formed color of image feature description vectors and image, and images all in database is completed to processing.Processed offline module is responsible for the image in database to process, for the every piece image in database produces corresponding color feature vector, by color feature vector sum image together corresponding stored in database.While carrying out follow-up feature comparison, can directly import the color feature vector of image, without image is processed online, to improve the execution speed of system.
Online processing module 22 is responsible for obtaining query image from terminal, after being processed, query image forms color feature vector, compare with the color feature vector of image in database, from database, choose the image the highest with query image color similarity, and net result is shown to terminal.Line processing module 22 workflows comprise that query image obtains 211b, color of image space conversion 212b, color fuzzy quantization and process 213b, color feature 214b, feature relatively 215 and extract result and show 216, are described in detail as follows:
Step 211b, query image is obtained.From terminal, obtain image, read in online processing module.The image reading in is for example used RGB color space to represent in system.
Step 212b, the conversion of color of image space, is transformed into HSV space by rgb space.After color of image space converts, each pixel of image is used tone h, saturation degree s and brightness v to represent, wherein the span of h is that 0 to 360, s span is that 0 to 1, v span is 0 to 1.
Step 213b, color fuzzy quantization is processed.According to the color attribute in hsv color space, tone h component is carried out to fuzzy quantization, s and v component adopt the mode directly quantizing to process.After image is carried out to the processing of color fuzzy quantization, after each pixel quantizes, the span of color value is between 0 ~ N-1.
Step 214b, color feature.Color value in statistic quantification image is respectively 0 to N-1 pixel quantity, and is normalized with the sum of all pixels of image, forms the query image color feature vector of N dimension.
Step 215, feature comparison.Read in the database images color feature vector file generating in off-line module, use the query image color feature vector obtaining to compare with described file, adopt absolute value distance as the standard of calculating correlativity, calculate the color correlation of every piece image in query image and database.
Step 216, extracts color associated picture, carries out result demonstration.Color correlation to every piece image in the query image obtaining in step 215 and database sorts from big to small, chooses former width images that correlativity is the highest, and in database images color feature vector file, obtains the title of these images.According to title, from database, transfer corresponding image and send to terminal and extract result demonstration.
In above-described embodiment, system is processed the image in database images by processed offline module, for the every piece image in database generates corresponding color feature vector, and the title of this vector sum image is saved in database images color feature vector file together.When using query image to carry out color associated picture extraction operation, directly reading data storehouse color of image feature description vectors file carries out Characteristic Contrast, without database images is processed online, can effectively improve the execution speed of system.
In one embodiment, in hsv color space, by Color Image Quantization, be 36 kinds of color values, and utilize fuzzy control to eliminate the quantization error being caused by border uncontinuity, color histogram after statistical picture quantizes extracts the image relevant to query image color from database as color characteristic.
Below in conjunction with Fig. 3 to Fig. 6, it is example introduction key step wherein.
Fig. 3 illustrates the process flow diagram of an example of color space conversion of the present invention, and the key step of conversion is as follows:
Step 3b, the conversion of saturation degree s.In image, to each pixel calculate respectively its R, G, tri-channel value of B and, choose and maximal value as the normalized parameter of saturation degree.When calculating the saturation degree s of single pixel, choose maximal value and minimum value in this pixel R, G, tri-channel value of B, their official post is carried out to normalizing calculating with above-mentioned normalized parameter, can obtain the value of saturation degree s in HSV space, its span is between 0 to 1.
It may be noted that step 3a, 3b and 3c can sequentially carry out or parallel processing, the order between each step is not restricted yet.
The color of coloured image is enriched very much, and a width rgb image just can have the color that 256*256*256 kind is different, and while directly carrying out color comparison, calculated amount is very large.Consider that human eye is limited to the resolution of color, within the scope of certain, human eye even cannot be perceiveed the variation of color is insensitive, can use the mode of color quantizing to reduce color sum, thereby reduce the dimension of color characteristic, shorten the computing time of color comparison.Generally speaking, in visual signature, the difference between color is mainly embodied by the difference of shade of color.For reducing computation complexity, when carrying out color quantizing, only tone h component is carried out to fuzzy quantization, s and v component adopt the mode directly quantizing to process.
Fig. 4 illustrates color quantizing main flow chart in one embodiment of the present of invention, and the key step of color quantizing is as follows:
As shown in Figure 5, step 501, luminance quantization.If the value of brightness v is between 0.2 and 0.7, its quantized value is 0; If the value of brightness v is between 0.7 and 1, its quantized value is 1.
I=4H+2S+V+8 （1）
Wherein, I is quantized color value, the quantized value that H is tone, the quantized value that S is saturation degree, the quantized value that V is brightness.
Fig. 6 has introduced the process flow diagram of an example of tone fuzzy quantization, in this example, and quantized color value value 0 ~ 35.Concrete step is as follows:
Does step 604, judge that the tone h value of pixel drops in zone of transition? if so,, continue step 605, otherwise, step 606 continued.
After image is carried out to the processing of color fuzzy quantization, after each pixel quantizes, the span of color value is between 0 ~ 35.Color value in statistic quantification image is respectively 0 to 35 pixel quantity, and is normalized with the sum of all pixels of image, has just formed 36 dimension color feature vectors of this image.
A kind of method of using color characteristic that fuzzy quantization was processed to carry out associated picture extraction is provided in above-described embodiment, overcome the shortcoming and defect of the image retrieval technologies based on text marking, for helping people to extract the image relevant to query image color from large capacity image library fast and accurately.
Fig. 7 illustrates the structural drawing of an embodiment of image indexing system of the present invention.As shown in Figure 7, this image indexing system comprises: color space represents acquisition module 71, for obtaining the hsv color space representation of tone, saturation degree and the brightness of image pixel; Color space represents quantization modules 72, for the hsv color space representation to image pixel, carries out quantification treatment, obtains and belongs to the quantized color value within the scope of 0 ~ N, and N is natural number, wherein, tone value is carried out to non-homogeneous fuzzy quantization; Color feature vector acquisition module 73, the histogram for the quantized color value of statistical picture pixel, is normalized with the sum of all pixels of image, obtains the color feature vector of image; Color feature vector comparison module 74, for comparing the color feature vector of the color feature vector of image and other images to determine the similarity of image.In one embodiment, color space represents acquisition module, and the pixel of image is represented to be converted to hsv color space representation by RGB color space, wherein, the span of tone is 0 to 360, and the span of saturation degree is 0 to 1, and the span of brightness is 0 to 1.
Fig. 8 illustrates the structural drawing of another embodiment of image indexing system of the present invention.In this embodiment, color space represents that quantization modules 82 comprises: luminance quantization unit 822, for brightness value being quantized to obtain luminance quantization value; Saturation degree quantifying unit 823, for quantizing to obtain saturation degree quantized value to intensity value; Tone quantifying unit 824, obtains tone quantized value for tone value being carried out to non-homogeneous fuzzy quantization; Quantized color value determining unit 821, for determining the quantized color value that belongs to the pixel within the scope of 0 ~ N according to the brightness value of pixel, intensity value and luminance quantization value, saturation degree quantized value and tone quantized value.
In one embodiment, tone quantifying unit arranges zone of transition at the intersection of tone, if tone value drops in zone of transition, tone value is quantified as to the weighted sum of the color quantization value relevant to zone of transition, if tone value does not drop in zone of transition, tone value is directly quantized.For example, tone quantifying unit is for arranging a plurality of quantization boundaries of tone; On the both sides of quantization boundary, quantization boundary zone of transition is set, the size of zone of transition and the range size between two adjacent quantization boundaries are directly proportional; Obtain the membership function of zone of transition, according to ridge type, distribute, try to achieve the membership function value that tone value belongs to zone of transition; If the tone value of pixel drops in zone of transition, the tone value of pixel is quantified as to the weighted sum of two the color quantization values relevant to this zone of transition, weighted value is membership function value; If the tone value of pixel drops on outside zone of transition, directly according to tone value, determine corresponding quantized value.In one embodiment, N is 36, and quantization boundary comprises 22,45,70,155,186,278,330.
In one embodiment, if the brightness value of quantized color value determining unit judgement pixel is less than 0.2, the quantized color value value of pixel is 0; If the brightness value of pixel is more than or equal to 0.8, the value of saturation degree is less than 0.2, and the quantized color value value of pixel is 7; If the brightness value of pixel is between 0.2 and 0.8, the value of saturation degree is less than 0.2, the value after 10 demultiplications 1 that the quantized color value value of pixel is brightness value also round downwards; If the brightness value of pixel is more than or equal to 0.2, the value of saturation degree s is more than or equal to 0.2, and the quantized color value of pixel is obtained the quantized color value of pixel by the weighting of the tone quantized value of luminance quantization value, saturation degree quantifying unit acquisition saturation degree quantized value and the acquisition of tone quantifying unit of luminance quantization unit acquisition pixel.If the brightness value of pixel is more than or equal to 0.2, the value of saturation degree is more than or equal to 0.2, if the brightness value of luminance quantization unit judges pixel is between 0.2 and 0.7, the quantized value of the brightness value of pixel is 0, if the value of brightness v is between 0.7 and 1, the quantized value of the brightness value of pixel is 1; If the intensity value of saturation degree quantifying unit judgement pixel is between 0.2 and 0.65, the quantized value of the intensity value of pixel is 0, if the intensity value of pixel between 0.65 and 1, the quantized value of the intensity value of pixel is 1; Tone quantifying unit is carried out non-homogeneous fuzzy quantization processing to the tone value of pixel.
The color feature vector that embodiments of the invention are usingd after image quantization, as relevant basis for estimation, extracts the image relevant to query image color in database, makes image indexing system more intelligent, and the extraction result of color associated picture is also more accurate.By adopting off-line module to process in advance the image in database images, effectively improved execution speed.
So far, described in detail according to image search method of the present invention and system.For fear of covering design of the present invention, details more known in the field are not described.Those skilled in the art, according to description above, can understand how to implement technical scheme disclosed herein completely.
May realize in many ways method and system of the present invention.For example, can realize method and system of the present invention by any combination of software, hardware, firmware or software, hardware, firmware.The said sequence that is used for the step of method is only in order to describe, and the step of method of the present invention is not limited to above specifically described order, unless otherwise specified.In addition, in certain embodiments, can be also the program being recorded in recording medium by the invention process, these programs comprise for realizing the machine readable instructions of the method according to this invention.Thereby the present invention also covers storage for carrying out the recording medium of the program of the method according to this invention.
Although specific embodiments more of the present invention are had been described in detail by example, it should be appreciated by those skilled in the art, above example is only in order to describe, rather than in order to limit the scope of the invention.It should be appreciated by those skilled in the art, can without departing from the scope and spirit of the present invention, above embodiment be modified.Scope of the present invention is limited by claims.
Claims (17)
1. an image search method, is characterized in that, comprising:
Obtain tone, saturation degree and the brightness hsv color space representation of pixel in image;
Hsv color space representation to pixel in described image carries out quantification treatment, obtains and belongs to the quantized color value within the scope of 0 ~ N, and N is natural number, wherein, tone value is carried out to non-homogeneous fuzzy quantization;
The histogram of adding up the quantized color value of described image pixel, is normalized with the sum of all pixels of described image, obtains the color feature vector of described image;
The color feature vector of the color feature vector of described image and other images is compared to determine to the similarity of image.
2. method according to claim 1, is characterized in that, describedly tone value is carried out to non-homogeneous fuzzy quantization comprises:
Intersection at tone arranges zone of transition, if tone value drops in described zone of transition, described tone value is quantified as to the weighted sum of the color quantization value relevant to described zone of transition, if described tone value does not drop in described zone of transition, described tone value is directly quantized.
3. method according to claim 1, is characterized in that, describedly tone value is carried out to non-homogeneous fuzzy quantization comprises:
A plurality of quantization boundaries of tone are set;
On the both sides of described quantization boundary, quantization boundary zone of transition is set, the size of described zone of transition and the range size between two adjacent quantization boundaries are directly proportional;
Obtain the membership function of described zone of transition, according to ridge type, distribute, try to achieve the membership function value that tone value belongs to zone of transition;
If the tone value of pixel drops in described zone of transition, the tone value of described pixel is quantified as to the weighted sum of two the color quantization values relevant to described zone of transition, weighted value is membership function value;
If the tone value of pixel drops on outside zone of transition, according to tone value, determine corresponding quantized value.
4. method according to claim 3, is characterized in that, described N is 36, and described quantization boundary comprises 22,45,70,155,186,278,330.
5. method according to claim 1, is characterized in that, also comprises:
The pixel of described image is represented to be converted to hsv color space representation by RGB RGB color space, and wherein, the span of tone is 0 to 360, and the span of saturation degree is 0 to 1, and the span of brightness is 0 to 1.
6. method according to claim 5, is characterized in that, the described hsv color space representation to pixel in described image carries out quantification treatment and obtains the quantized color value belong within the scope of 0 ~ N and comprise:
If the brightness value of pixel is less than 0.2, the quantized color value value of pixel is 0;
If the brightness value of pixel is more than or equal to 0.8, the value of saturation degree is less than 0.2, and the quantized color value value of pixel is 7;
If the brightness value of pixel is between 0.2 and 0.8, the value of saturation degree is less than 0.2, the value after 10 demultiplications 1 that the quantized color value value of pixel is brightness value also round downwards;
If the brightness value of pixel is more than or equal to 0.2, the value of saturation degree is more than or equal to 0.2, and the quantized color value of pixel is obtained by the quantized value weighting of brightness value, intensity value and the tone value of pixel.
7. method according to claim 6, it is characterized in that, if the brightness value of described pixel is more than or equal to 0.2, the value of saturation degree is more than or equal to 0.2, the quantized color value of pixel is obtained and comprised by the quantized value weighting of brightness value, intensity value and the tone value of pixel:
If the brightness value of pixel is between 0.2 and 0.7, the quantized value of the brightness value of pixel is 0; If the value of brightness v is between 0.7 and 1, the quantized value of the brightness value of pixel is 1;
If the intensity value of pixel is between 0.2 and 0.65, the quantized value of the intensity value of pixel is 0; If the intensity value of pixel is between 0.65 and 1, the quantized value of the intensity value of pixel is 1;
The tone value of pixel is carried out to non-homogeneous fuzzy quantization processing;
The quantized color value of pixel is obtained by the quantized value weighting of tone value, intensity value and the brightness value of pixel:
I=4H+2S+V+8
Wherein, I is quantized color value, the quantized value that H is tone, the quantized value that S is saturation degree, the quantized value that V is brightness.
8. method according to claim 1, is characterized in that, the described color feature vector by the color feature vector of described image and other images compares to determine that similar image comprises:
The color feature vector of the color feature vector of described image and other images is compared, according to the absolute value distance between color feature vector or Euclidean distance, calculate the similarity of described image and other images.
9. an image indexing system, is characterized in that, comprising:
Color space represents acquisition module, for obtaining tone, saturation degree and the brightness hsv color space representation of image pixel;
Color space represents quantization modules, for the hsv color space representation to described image pixel, carries out quantification treatment, obtains and belongs to the quantized color value within the scope of 0 ~ N, and N is natural number, wherein, tone value is carried out to non-homogeneous fuzzy quantization;
Color feature vector acquisition module, for adding up the histogram of the quantized color value of described image pixel, is normalized with the sum of all pixels of described image, obtains the color feature vector of described image;
Color feature vector comparison module, for comparing the color feature vector of the color feature vector of described image and other images to determine the similarity of image.
10. system according to claim 9, is characterized in that, described color space represents that quantization modules comprises:
Luminance quantization unit, for quantizing to obtain luminance quantization value to brightness value;
Saturation degree quantifying unit, for quantizing to obtain saturation degree quantized value to intensity value;
Tone quantifying unit, obtains tone quantized value for tone value being carried out to non-homogeneous fuzzy quantization;
Quantized color value determining unit, for determining the quantized color value that belongs to the pixel within the scope of 0 ~ N according to the brightness value of pixel, intensity value and described luminance quantization value, saturation degree quantized value and tone quantized value.
11. systems according to claim 10, it is characterized in that, described tone quantifying unit arranges zone of transition at the intersection of tone, if tone value drops in described zone of transition, described tone value is quantified as to the weighted sum of the color quantization value relevant to described zone of transition, if described tone value does not drop in described zone of transition, described tone value is directly quantized.
12. systems according to claim 10, is characterized in that, described tone quantifying unit is for arranging a plurality of quantization boundaries of tone; On the both sides of described quantization boundary, quantization boundary zone of transition is set, the size of described zone of transition and the range size between two adjacent quantization boundaries are directly proportional; Obtain the membership function of described zone of transition, according to ridge type, distribute, try to achieve the membership function value that tone value belongs to zone of transition; If the tone value of pixel drops in described zone of transition, the tone value of described pixel is quantified as to the weighted sum of two the color quantization values relevant to this zone of transition, weighted value is membership function value; If the tone value of pixel drops on outside zone of transition, directly according to tone value, determine corresponding quantized value.
13. systems according to claim 12, is characterized in that, described N is 36, and described quantization boundary comprises 22,45,70,155,186,278,330.
14. systems according to claim 10, is characterized in that, if the brightness value of described quantized color value determining unit judgement pixel is less than 0.2, the quantized color value value of pixel is 0; If the brightness value of pixel is more than or equal to 0.8, the value of saturation degree is less than 0.2, and the quantized color value value of pixel is 7; If the brightness value of pixel is between 0.2 and 0.8, the value of saturation degree is less than 0.2, the value after 10 demultiplications 1 that the quantized color value value of pixel is brightness value also round downwards; If the brightness value of pixel is more than or equal to 0.2, the value of saturation degree s is more than or equal to 0.2, and the weighting of the tone quantized value that the quantized color value of pixel is obtained by the luminance quantization value of described luminance quantization unit acquisition pixel, described saturation degree quantifying unit acquisition saturation degree quantized value and described tone quantifying unit obtains the quantized color value of pixel.
15. systems according to claim 14, it is characterized in that, if the brightness value of pixel be more than or equal to 0.2 and the value of saturation degree be more than or equal to 0.2, if the brightness value of described luminance quantization unit judges pixel is between 0.2 and 0.7, the quantized value of the brightness value of pixel is 0, if the value of brightness v is between 0.7 and 1, the quantized value of the brightness value of pixel is 1; If the intensity value of described saturation degree quantifying unit judgement pixel is between 0.2 and 0.65, the quantized value of the intensity value of pixel is 0, if the intensity value of pixel between 0.65 and 1, the quantized value of the intensity value of pixel is 1; Described tone quantifying unit is carried out non-homogeneous fuzzy quantization processing to the tone value of pixel.
16. systems according to claim 9, it is characterized in that, described color space represents acquisition module, the pixel of described image is represented to be converted to hsv color space representation by RGB RGB color space, wherein, the span of tone is 0 to 360, and the span of saturation degree is 0 to 1, and the span of brightness is 0 to 1.
17. systems according to claim 9, it is characterized in that, described color feature vector comparison module compares the color feature vector of the color feature vector of described image and other images, calculates the similarity of described image and other images according to the absolute value distance between color feature vector or Euclidean distance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210365623.XA CN103699532B (en) | 2012-09-27 | 2012-09-27 | Image color retrieval method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210365623.XA CN103699532B (en) | 2012-09-27 | 2012-09-27 | Image color retrieval method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103699532A true CN103699532A (en) | 2014-04-02 |
| CN103699532B CN103699532B (en) | 2017-08-25 |
Family
ID=50361063
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210365623.XA Active CN103699532B (en) | 2012-09-27 | 2012-09-27 | Image color retrieval method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103699532B (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105488510A (en) * | 2015-11-20 | 2016-04-13 | 上海华力创通半导体有限公司 | Construction method and system for color histogram of static picture |
| CN105740789A (en) * | 2016-01-26 | 2016-07-06 | 浙江捷尚视觉科技股份有限公司 | Video target retrieval method based on color features |
| CN105913460A (en) * | 2015-11-26 | 2016-08-31 | 乐视致新电子科技（天津）有限公司 | Skin color detection method and device |
| CN106599185A (en) * | 2016-12-14 | 2017-04-26 | 北京微智信业科技有限公司 | HSV-based image similarity identification method |
| CN107358635A (en) * | 2017-07-19 | 2017-11-17 | 辽宁工程技术大学 | A kind of Color-scale Morphology image processing method based on fuzzy comparability |
| CN108279238A (en) * | 2018-01-30 | 2018-07-13 | 深圳春沐源控股有限公司 | A kind of fruit maturity judgment method and device |
| CN108804475A (en) * | 2017-05-05 | 2018-11-13 | 北京京东尚科信息技术有限公司 | The method and apparatus for searching for color similar pictures |
| CN110533622A (en) * | 2019-08-27 | 2019-12-03 | 辽宁东智威视科技有限公司 | Parameter automatization configuration method in a kind of synthesis of picture |
| CN110647910A (en) * | 2019-08-12 | 2020-01-03 | 浙江浩腾电子科技股份有限公司 | Image similarity calculation method based on color quantization |
| CN110717444A (en) * | 2019-10-09 | 2020-01-21 | 北京明略软件系统有限公司 | Lipstick number identification method and device |
| CN110728722A (en) * | 2019-09-18 | 2020-01-24 | 苏宁云计算有限公司 | Image color migration method and device, computer equipment and storage medium |
| CN111787671A (en) * | 2020-07-15 | 2020-10-16 | 江门市征极光兆科技有限公司 | Control method based on movie and television picture synchronous light atmosphere |
| CN113096149A (en) * | 2019-12-23 | 2021-07-09 | 北矿机电科技有限责任公司 | Shaking table ore belt segmentation method based on three color elements |
| CN113127670A (en) * | 2019-12-31 | 2021-07-16 | 飞书数字科技(上海)有限公司 | Method, device, storage medium and processor for searching target color |
| CN114638826A (en) * | 2022-05-13 | 2022-06-17 | 河南银金达新材料股份有限公司 | Method for detecting optical fatigue degree of photochromic barrier film |
| CN115830352A (en) * | 2023-02-20 | 2023-03-21 | 深圳中微电科技有限公司 | Image similarity contrast method and device and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090087016A1 (en) * | 2007-09-28 | 2009-04-02 | Alexander Berestov | Content based adjustment of an image |
| CN101763429A (en) * | 2010-01-14 | 2010-06-30 | 中山大学 | Image retrieval method based on color and shape features |
-
2012
- 2012-09-27 CN CN201210365623.XA patent/CN103699532B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090087016A1 (en) * | 2007-09-28 | 2009-04-02 | Alexander Berestov | Content based adjustment of an image |
| CN101763429A (en) * | 2010-01-14 | 2010-06-30 | 中山大学 | Image retrieval method based on color and shape features |
Non-Patent Citations (2)
| Title |
|---|
| 于仕: ""基于颜色特征的图像检索技术研究"", 《中国优秀硕士学位论文全文数据库（信息科技辑）》 * |
| 鲍锡义，等: ""一种色调变化模糊量化方法"", 《西安科技大学学报》 * |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105488510A (en) * | 2015-11-20 | 2016-04-13 | 上海华力创通半导体有限公司 | Construction method and system for color histogram of static picture |
| CN105488510B (en) * | 2015-11-20 | 2019-01-15 | 上海华力创通半导体有限公司 | The construction method and its system of the color histogram of static images |
| CN105913460A (en) * | 2015-11-26 | 2016-08-31 | 乐视致新电子科技（天津）有限公司 | Skin color detection method and device |
| CN105740789A (en) * | 2016-01-26 | 2016-07-06 | 浙江捷尚视觉科技股份有限公司 | Video target retrieval method based on color features |
| CN105740789B (en) * | 2016-01-26 | 2018-12-07 | 浙江捷尚视觉科技股份有限公司 | A kind of video object search method based on color characteristic |
| CN106599185B (en) * | 2016-12-14 | 2020-10-23 | 北京微智信业科技有限公司 | HSV-based image similarity identification method |
| CN106599185A (en) * | 2016-12-14 | 2017-04-26 | 北京微智信业科技有限公司 | HSV-based image similarity identification method |
| CN108804475A (en) * | 2017-05-05 | 2018-11-13 | 北京京东尚科信息技术有限公司 | The method and apparatus for searching for color similar pictures |
| CN107358635A (en) * | 2017-07-19 | 2017-11-17 | 辽宁工程技术大学 | A kind of Color-scale Morphology image processing method based on fuzzy comparability |
| CN107358635B (en) * | 2017-07-19 | 2020-11-03 | 辽宁工程技术大学 | Color morphological image processing method based on fuzzy similarity |
| CN108279238A (en) * | 2018-01-30 | 2018-07-13 | 深圳春沐源控股有限公司 | A kind of fruit maturity judgment method and device |
| CN110647910A (en) * | 2019-08-12 | 2020-01-03 | 浙江浩腾电子科技股份有限公司 | Image similarity calculation method based on color quantization |
| CN110533622A (en) * | 2019-08-27 | 2019-12-03 | 辽宁东智威视科技有限公司 | Parameter automatization configuration method in a kind of synthesis of picture |
| CN110533622B (en) * | 2019-08-27 | 2022-03-25 | 辽宁东智威视科技有限公司 | Automatic parameter configuration method in picture synthesis |
| CN110728722A (en) * | 2019-09-18 | 2020-01-24 | 苏宁云计算有限公司 | Image color migration method and device, computer equipment and storage medium |
| CN110728722B (en) * | 2019-09-18 | 2022-08-02 | 苏宁云计算有限公司 | Image color migration method and device, computer equipment and storage medium |
| CN110717444A (en) * | 2019-10-09 | 2020-01-21 | 北京明略软件系统有限公司 | Lipstick number identification method and device |
| CN113096149A (en) * | 2019-12-23 | 2021-07-09 | 北矿机电科技有限责任公司 | Shaking table ore belt segmentation method based on three color elements |
| CN113096149B (en) * | 2019-12-23 | 2023-11-10 | 北矿机电科技有限责任公司 | Shaking table ore belt segmentation method based on three color elements |
| CN113127670A (en) * | 2019-12-31 | 2021-07-16 | 飞书数字科技(上海)有限公司 | Method, device, storage medium and processor for searching target color |
| CN111787671A (en) * | 2020-07-15 | 2020-10-16 | 江门市征极光兆科技有限公司 | Control method based on movie and television picture synchronous light atmosphere |
| CN114638826A (en) * | 2022-05-13 | 2022-06-17 | 河南银金达新材料股份有限公司 | Method for detecting optical fatigue degree of photochromic barrier film |
| CN115830352A (en) * | 2023-02-20 | 2023-03-21 | 深圳中微电科技有限公司 | Image similarity contrast method and device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103699532B (en) | 2017-08-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103699532A (en) | Image color retrieval method and system | |
| WO2018028583A1 (en) | Subtitle extraction method and device, and storage medium | |
| CN102509118B (en) | Method for monitoring video retrieval | |
| CN102176208B (en) | Robust video fingerprint method based on three-dimensional space-time characteristics | |
| Küçüktunç et al. | Fuzzy color histogram-based video segmentation | |
| US20130258198A1 (en) | Video search system and method | |
| US9596520B2 (en) | Method and system for pushing information to a client | |
| US9075876B2 (en) | Search method for video clip | |
| CN103632153B (en) | Region-based image saliency map extracting method | |
| CN101388020A (en) | Composite image search method based on content | |
| JP4506409B2 (en) | Region dividing method and apparatus, image recognition processing apparatus, program, and recording medium | |
| CN104410867A (en) | Improved video shot detection method | |
| CN114926826A (en) | Scene text detection system | |
| KR20100071822A (en) | Apparatus and method for detecting clothes in image | |
| Yang et al. | No‐reference image quality assessment via structural information fluctuation | |
| Zheng et al. | Study on image retrieval based on image texture and color statistical projection | |
| CN111898544B (en) | Text image matching method, device and equipment and computer storage medium | |
| Kong et al. | Full-reference IPTV image quality assessment by deeply learning structural cues | |
| Moghimi et al. | Shadow detection based on combinations of hessenberg decomposition and principal component analysis in surveillance applications | |
| Lizarraga-Morales et al. | Integration of color and texture cues in a rough set–based segmentation method | |
| US20150026013A1 (en) | System and methods for cognitive visual product search | |
| Glotin et al. | Efficient image concept indexing by harmonic & arithmetic profiles entropy | |
| Sekhar et al. | An object-based splicing forgery detection using multiple noise features | |
| KR20150073409A (en) | Apparatus and method for near duplicate video clip detection | |
| Yang et al. | Image perception hash and structural similarity fusion model for parking lot status recognition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |