CN106469437A - Image processing method and image processing apparatus - Google Patents
Image processing method and image processing apparatus Download PDFInfo
- Publication number
- CN106469437A CN106469437A CN201510508531.6A CN201510508531A CN106469437A CN 106469437 A CN106469437 A CN 106469437A CN 201510508531 A CN201510508531 A CN 201510508531A CN 106469437 A CN106469437 A CN 106469437A
- Authority
- CN
- China
- Prior art keywords
- image block
- described image
- description information
- semantic description
- image
- 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
Landscapes
- Image Analysis (AREA)
Abstract
Provide a kind of image processing method and image processing apparatus.Described image processing method includes：Pending image averaging is divided into multiple images block；Obtain the semantic description information of described image block；Determine the semantic description information of described image based on the semantic description information of locus in described image for the described image block and described image block.In the technical scheme of the embodiment of the present disclosure, can organically contact associated picture block in the whole image semantic description information to determine image, more meet the understanding custom to image.
Description
Technical field
The present invention relates to areas of information technology, at a kind of image processing method and image
Reason device.
Background technology
Image understanding (image understanding, IU) is the semantic understanding to image.Image understanding is
With image as object, knowledge is core, the target in research image, the mutual relation between target, figure
The scene of picture and its application.
Complete picture material, as the basic description carrier of knowledge information, can be changed by semantic description information
One-tenth can intuitivism apprehension class text language performance, image understanding plays vital effect.Image
In abundant semantic description information more accurate image search engine can be provided, generate the digitized map of intelligence
As the visual scene description in photograph album and virtual world.
As the mode of the semantic description information generating image, the visual signature being typically based on image is to image
Split and obtained multiple images region, each image-region after segmentation is separate, then
Analyze the semantic description information of each image-region, and the semantic description information based on each image-region Lai
Obtain the semantic description information of whole image.
Content of the invention
The embodiment of the present disclosure provides a kind of image processing method and image processing apparatus, and it provides one kind
The image procossing mode of the semantic description information of new determination image, this image procossing mode organically contacts
Related image block, more meets the understanding custom to image.
A kind of first aspect, there is provided image processing method.This image processing method may include：To wait to locate
The image averaging of reason is divided into multiple images block；Obtain the semantic description information of described image block；Based on institute
State locus in described image for the image block and the semantic description information of described image block determines described figure
The semantic description information of picture.
In conjunction with a first aspect, in a kind of implementation of first aspect, the described described image block that obtains
Semantic description information may include：Obtain the Gaussian Mixture with each corresponding picture material of semantic description information
Model；Determine the semantic description information of described image block according to described image block and described gauss hybrid models.
In conjunction with first aspect and its above-mentioned implementation, in another implementation of first aspect, described
Can be wrapped according to the semantic description information that described image block and described gauss hybrid models determine described image block
Include：Determine that described image block belongs to institute according to the similarity of described image block and described gauss hybrid models
State the probability Estimation of each picture material；The probability of each picture material according to described image block belongs to
Estimate to determine the semantic description information of described image block.
In conjunction with first aspect and its above-mentioned implementation, in another implementation of first aspect, described
Semantic description information based on locus in described image for the described image block and described image block determines
The semantic description information of described image may include：Semantic description information according to described image block and described figure
As locus in described image for the block determine the weight of described image block；Language according to described image block
The weight of adopted description information and described image block determines the semantic description information of described image.
In conjunction with first aspect and its above-mentioned implementation, in another implementation of first aspect, described
Locus in described image for the semantic description information and described image block according to described image block determine
The weight of described image block may include：Semantic description information according to described image block calculates described image block
It is adjacent the similarity between image block；It is adjacent similar between image block based on described image block
The weight of degree setting described image block.
In conjunction with first aspect and its above-mentioned implementation, in another implementation of first aspect, described
Semantic description information based on locus in described image for the described image block and described image block determines
The semantic description information of described image may include：Based on locus in described image for the described image block
Determine the adjacent image block conduct with similar semantic description information with the semantic description information of described image block
Content aggregation region, using described similar semantic description information as described Content aggregation region semantic description
Information；Semantic description information according to described Content aggregation region and described Content aggregation region are in described figure
Locus in picture determine the weight in described Content aggregation region；Power according to described Content aggregation region
The semantic description information in weight and described Content aggregation region determines the semantic description information of described image.
A kind of second aspect, there is provided image processing apparatus.This image processing apparatus may include：Memorizer；
And processor, for executing following operation：Pending image averaging is divided into multiple images block；Obtain
Obtain the semantic description information of described image block；Based on locus in described image for the described image block and
The semantic description information of described image block determines the semantic description information of described image.
A kind of third aspect, there is provided image processing apparatus.This image processing apparatus may include division unit
910th, image block semanteme determining unit 920 and image, semantic determining unit 930.Division unit 910 will be treated
The image averaging processing is divided into multiple images block.Image block semanteme determining unit 920 obtains described image
The semantic description information of block.Image, semantic determining unit 930 is based on described image block in described image
The semantic description information of locus and described image block determines the semantic description information of described image
In the technical scheme of the image processing method according to the embodiment of the present disclosure and image processing apparatus, lead to
Cross the locus in whole image based on each image block true with the semantic description information of each image block
Determine the semantic description information of whole image, thus organically contacted by the associated picture block in whole image,
More meet the understanding custom to image.
Brief description
In order to be illustrated more clearly that the technical scheme of the embodiment of the present disclosure, below will be to embodiment or existing skill
Art description in required use accompanying drawing be briefly described it should be apparent that, drawings in the following description
It is only some embodiments of the present disclosure, for those of ordinary skill in the art, can also be according to this
A little accompanying drawings obtain other accompanying drawings.
Fig. 1 is the flow chart schematically illustrating the image processing method according to the embodiment of the present disclosure.
Fig. 2 schematically illustrates the example of the image being divided into multiple images block.
Fig. 3 is that the semanteme of each image block of acquisition in the image processing method schematically illustrate Fig. 1 is retouched
The flow chart stating information.
Fig. 4 schematically illustrates the semantic description information of obtained image block.
Fig. 5 be a diagram that the first of the semantic description information of the determination image in the image processing method of Fig. 1
The flow chart of example.
Fig. 6 schematically illustrate determined by image semantic description information.
Fig. 7 be a diagram that the second of the semantic description information of the determination image in the image processing method of Fig. 1
The flow chart of example.
Fig. 8 is the block diagram schematically illustrating the first image processing apparatus according to the embodiment of the present disclosure.
Fig. 9 is the block diagram schematically illustrating the second image processing apparatus according to the embodiment of the present disclosure.
Specific embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present disclosure, the technical scheme in the embodiment of the present disclosure is carried out
Clearly and completely description is it is clear that described embodiment a part of embodiment that is the disclosure, rather than
Whole embodiments.
For an image, machine generally cannot understand its content, thus being difficult to picture search.At this
In disclosed embodiment, pending image carried out process with can convert images into can intuitivism apprehension
Semantic description information, can provide more accurate image search engine based on this semantic description information, thus raw
Become the visual scene description in the digital picture photograph album and virtual world of intelligence.Additionally, the language based on image
Adopted description information, can also carry out image labeling, image recognition etc..Described pending image can be
From web search to image, can be the image using collections such as photographing units.The acquisition of pending image
Mode does not constitute the restriction to the embodiment of the present disclosure.
Fig. 1 is the flow chart schematically illustrating the image processing method 100 according to the embodiment of the present disclosure.
As shown in figure 1, this image processing method 100 includes：Pending image averaging is divided into multiple figures
As block (S110)；Obtain the semantic description information (S120) of described image block；Existed based on described image block
The semantic description information of the locus in described image and described image block determines that the semanteme of described image is retouched
State information (S130).
In S110, pending image averaging is divided into multiple images block.Pending image is usual
To be represented with the numerical value of each pixel of image.For each pixel, can be divided with gray value, three primary colories
Amount etc. represents.Image typically comprises the picture element matrix arranging according to row and column mode.Many as being divided into
The example of individual image block, can divide successively according to the mode that each image block includes 16 × 16 pixels.
Fig. 2 schematically illustrates the example of the pending image being divided into multiple images block.This waits to locate
The image of reason is a seashore picture with scenes.It is light blue sky above this seashore picture with scenes, should
The middle part of seashore picture with scenes is azure sea, is golden sandy beach in the lower section of this seashore picture with scenes.
Assume that seashore picture with scenes has 128 × 112 pixels, if each image block B (1,8) includes 16 × 16
Pixel, then, from the beginning of the end of the picture element matrix of image, be expert at and column direction with 16 × 16 pixels for unit
Divide, this seashore picture with scenes can be divided into every 128/16=8 image block of behavior, often be classified as 112/16
56 (8 being multiplied by 7) the individual image block altogether of=7 image blocks.As shown in the grid lines in Fig. 2.
Assume that each image block B (i, j) represents, wherein i is line number in whole image for the image block B, j
It is columns in whole image for the image block B.
So that the total pixel number of pending image is divided equally just as a example it is illustrated in Fig. 2.Waiting to locate
In the case of the total pixel number of the image of reason is not the integral multiple of image block, it is also the end from picture element matrix
Start, be expert at 16 × 16 pixels for unit and column direction divides, and by last less than 16 × 16 pictures
The image-region of element is also divided into image block.
According in the embodiment of the present disclosure, pending image averaging is divided into multiple images block, this
Dividing mode is simple.However, in the image procossing mode of other semantic description information generating image,
The visual signature based on image may be needed image to be split and is obtained multiple images region, this is very
Complicated.
In S120, for each image block in the multiple images block dividing in S110, all obtain
The semantic description information of this image block.The implication of data is exactly semantic, and semantic description information is used to describe
Semantic information.View data inherently symbol, the data being only endowed implication can be used,
At this time data has translated into semantic description information.This semantic description information is, for example, one kind can be by machine
The class text language performance of device intuitivism apprehension.Taking Fig. 2 as a example, each image of described seashore picture with scenes
The semantic description information of block can be each image block is the general of at least one of sky, ocean, sandy beach
Rate is distributed.
Fig. 3 is that the semanteme of each image block of acquisition in the image processing method schematically illustrate Fig. 1 is retouched
The flow chart stating information S120.As shown in figure 3, obtaining each image block in the plurality of image block
Semantic description information includes：Obtain the Gaussian Mixture mould with each corresponding picture material of semantic description information
Type (S121)；Described image block is determined according to the similarity of described image block and described gauss hybrid models
Belong to the probability Estimation (S122) of each picture material described；Each figure according to described image block belongs to
As the probability Estimation of content determines the semantic description information (S123) of described image block.
In S121, each image corresponding with each semantic description information can be obtained from data base in advance
The training image of content, and the training image based on each picture material is obtaining the Gauss of this picture material
Mixed model.For example, the training image of each picture material, described image are prestored in data base
Content for example includes sky, ocean, sandy beach, meadow, great Shu, high mountain etc..Described image content is each
Plant involved content in the semantic description information of image.
Each picture material is likely to be of different images, for example, color under the conditions of different weather for the sky
Color and brightness is typically different.Therefore, accordingly it is likely to be of multiple training figures with each picture material
Picture, such as 512.For each picture material, a Gauss can be set up based on the plurality of training image
Model is characterizing this picture material.Gauss model utilizes normal distribution curve accurately quantized image content,
Gray scale in picture material, color etc. are decomposed into some models being formed based on normal distribution curve.
For each picture material, corresponding Gauss model is weighted averagely can be obtained by the height of this picture material
This mixed model.The mode of the acquisition gauss hybrid models taken does not constitute the limit to the embodiment of the present disclosure
System.Generally, the training image of each picture material and the Gaussian Mixture mould of each picture material are obtained ahead of time
Type, and set up data base.
In S122, according to each image block described and the described gauss hybrid models of each picture material
Similarity belongs to the probability Estimation of each picture material described determining each image block described.With in Fig. 2
The image block in the upper right corner as a example, by the view data of 16 × 16 pixels of image block B (1,8) respectively with
The gauss hybrid models of each picture material obtaining in S121 are mated.For example, by image block B (1,
8) view data respectively with the gauss hybrid models of sky, the gauss hybrid models of ocean, sandy beach height
This mixed model, the gauss hybrid models on meadow, the gauss hybrid models of big tree, the Gaussian Mixture of high mountain
Model is mated, to determine that image block B (1,8) belongs to sky, ocean, sandy beach, meadow, big respectively
Tree, the probability Estimation of high mountain.According to Fig. 2, the image block in the described upper right corner is sky, then pass through
The operation of S122 is it may be determined that image block B (1,8) belongs to the maximum probability of sky, and belongs to ocean, sand
Beach, meadow, great Shu, the probability of any one may very little even zero in high mountain.Here, suppose that image
Block B (1,8) belongs to sky, ocean, sandy beach, meadow, great Shu, the probability of high mountain are 90% respectively, 6%,
4%th, 0,0,0.
In S123, according to described image block belongs to, the probability Estimation of each picture material determines described figure
Semantic description information as block.For example, it is possible to represent that each image block belongs in each image with rectangular histogram
With ratio chart, the probability holding, can also represent that each image block belongs to the probability of each picture material.Depend on
In the specific requirement of described semantic description information, suitable treatments can be carried out to generate to described probability Estimation
Required semantic description information.
For each image block in pending image, all judge the institute of this image block and each picture material
State the similarity of gauss hybrid models and belong to the general of each picture material described determining each image block described
Rate is estimated, and determines its semantic description information based on described probability Estimation, thus obtaining pending figure
The semantic description information of each image block in picture.
In S122 and S123 in figure 3, determined according to described image block and described gauss hybrid models
The semantic description information of described image block, this is only example.Based on described gauss hybrid models, also may be used
To take other modes to determine the semantic description information of each image block.
Fig. 4 schematically illustrates the semantic description information of obtained image block.As shown in figure 4, for
The image block in the upper right corner in each image block shown in Fig. 2, the operation that execution combines Fig. 3 description is processed,
And obtain the semantic description information of the image block in the upper right corner representing with rectangular histogram.Diagram according to Fig. 4
As can be seen that the probability that image block B (1,8) belongs to sky is far longer than it and belongs to the probability at ocean or sandy beach,
Thus substantially can determine that image block B (1,8) belongs to sky.For convenience, illustrate only in Fig. 4
The semantic description information of one image block.
In fact, by the process of S120, having obtained the semantic description information of each image block in Fig. 2.
That is, having obtained and each image block one-to-one semantic description information.Continue taking Fig. 4 as a example,
In the diagram, basically illustrate the scenery at sky, ocean and sandy beach.In the image block showing sky,
Can obtain that there is the semantic description information that greater probability belongs to sky, the 1st row of such as Fig. 4 and the 2nd
The image block of row；In the image block showing ocean, can obtain there is greater probability and belong to ocean
3rd row of semantic description information, such as Fig. 4 and the image block of the 4th row；Showing the image at sandy beach
In block, can obtain that there is the semantic description information that greater probability belongs to sandy beach, the 6th row of such as Fig. 4
Image block with the 7th row.
Language in S130, based on locus in described image for the described image block and described image block
Adopted description information determines the semantic description information of described image.As an example, can be according to described image block
Locus in described image of semantic description information and described image block determine the power of described image block
Weight；The weight of the semantic description information according to described image block and described image block determines the language of described image
Adopted description information.By the locus in the picture according to image block, weight is set for this image block, can
The importance of each image block to distinguish is set, thus more accurately expressing the semantic description of whole image
Information.In other words, it is no longer separate between each image block dividing in S110, and be intended to
Determine the semantic description information of image based on each image block locus in the picture.
Fig. 5 be a diagram that the semantic description information (S130) of the determination image in the image processing method of Fig. 1
The first example flow chart.As shown in figure 5, institute is calculated according to the semantic description information of described image block
State the similarity (S131) that image block is adjacent between image block；Figure is adjacent based on described image block
As the similarity between block arranges the weight (S132) of described image block；Retouched according to the semanteme of described image block
The weight stating information and described image block determines the semantic description information (S133) of described image.
In S131, the semantic description information between two adjacent image blocks is closer to two neighbor map
As the similarity between block is bigger.Taking each image block in Fig. 4 as a example it is assumed that each image block only
It is related to three picture materials, i.e. sky, ocean and sandy beach.
For the image block B (Isosorbide-5-Nitrae) in Fig. 4, the image block being adjacent is image block B (1,3), image
Block B (1,5), image block B (2,4), the picture material of image block B (Isosorbide-5-Nitrae) and its adjacent image block is all with sky
Based on sky, ocean, the content at sandy beach occupy small part, so what this image block B (Isosorbide-5-Nitrae) was adjacent
Similarity between image block is larger.When calculating similarity, each picture material of image block to be considered
With the ratio occupied by each picture material.As an example, can be by calculating the semanteme of two image blocks
Euclidean distance between description information is calculating the similarity between this two image blocks.In the present embodiment,
Using four neighborhoods described in Fig. 4 as adjacent image block, it is possible to use eight neighborhood even more many neighborhoods conduct
Adjacent image block, does not limit here.
For the image block B (2,4) in Fig. 4, the image block being adjacent is image block B (2,3), image
Block B (2,5), image block B (1,4), image block B (3,4).Image block B (2,4), B (2,3), B (2,5),
In B (Isosorbide-5-Nitrae), all based on sky, ocean, the content at sandy beach occupy small part to picture material.However,
In image block B (3,4), based on ocean, sky, the content at sandy beach occupy small part to picture material.
Therefore, image block B (2,4) is high with the similarity of adjacent image block B (2,3), B (2,5), B (Isosorbide-5-Nitrae),
But low with the similarity of adjacent image block B (3,4), it reduce what image block B (2,4) was adjacent
The similarity of image block.Therefore, the similarity that image block B (2,4) is adjacent between image block is less than figure
As block B (1,4) is adjacent the similarity between image block.
In S132, it is adjacent the similarity setting described image between image block based on described image block
The weight of block.For example, in the case of the similarity height that image block is adjacent between image block, it is institute
State image block and high weight is set；In the case of the similarity that image block is adjacent between image block is low,
For described image block, low weight is set.For example, for above-mentioned image block B (Isosorbide-5-Nitrae) and B (2,4), image
Weight W (1,4) of block B (1,4) is more than weight W (2,4) of image block B (2,4).The like, thus
Obtain weight W (i, j) of each image block B (i, j) in Fig. 4.
In image shown in Fig. 4, in the first row of image and the second row based on sky, the 3rd
Row and fourth line in based on ocean, in fifth line before four row based on ocean, rear four row with sandy beach
Based on, in the 6th row and the 7th row based on sandy beach.When each image block picture material with about
The content of image block similarity higher, then the weight of this image block is bigger, more can represent figure exactly
As the semanteme in this region.
In S131 and S132, semantic description information according to described image block and described image block are in institute
State the weight that the locus in image determine described image block, this is only example, may also take on it
Its mode determines the weight of each image block, for example can be according to the Pixel Information of each image block Lai really
Fixed.
In S133, the semantic description information according to described image block and the weight of described image block determine institute
State the semantic description information of image.Multiple images content is potentially included, each picture material in whole image
Diverse location in image.Correspondingly, the semantic description information of image is the diverse location of image
The distribution of the semantic description information in region.
Continue, the language of each image block in the first row of image and the second row taking the image in Fig. 4 as a example
Adopted description information is all based on sky, and the ratio very little at ocean and sandy beach, and the first row each
Image block is very big with the similarity of surrounding image block, has very big weight, each image block in the second row
And three adjacent image blocks between, there is high similarity, only there is low phase seemingly with an adjacent image block
Degree, the weight of each image block in this second row is less than the weight of the image block in the first row as early as possible, but
It is that also there is higher weight, thus the semantic description information based on the first row and each image of the second row
Can determine that the semantic description information in this region for the image is sky with weight, and pole is occupied at ocean and sandy beach
Small scale.Similarly, using this S133 it may be determined that going out semantic description in the other positions region of image
Information.
Fig. 6 schematically illustrate determined by image semantic description information.As shown in the rightmost side of Fig. 6,
The method of the semantic description information of the determination image according to Fig. 5 has it can be determined that going out whole image
Three picture materials, i.e. sky, ocean and sandy beach.The first row of image and the second row can use the first language
Describing, the right side four row in the third line of image, fourth line and fifth line can be with for adopted description information
Describing, the left side four in the fifth line of image arranges two semantic description information, the 6th row, the 7th row can be used
3rd semantic description information is describing.First semantic description information is the ratio diagram of each picture material,
Wherein, sky occupies maximum ratio, and ocean and sandy beach occupy small percentage respectively.Second semantic description letter
Breath is the ratio diagram of each picture material, and wherein, ocean occupies maximum ratio, and sky and sandy beach are respectively
Occupy small percentage.3rd semantic description information is the ratio diagram of each picture material, wherein, sandy beach
Occupy maximum ratio, sky and ocean occupy small percentage respectively.
Mode above in conjunction with the semantic description information of the determination image of Fig. 5 and Fig. 6 description is only example.
In practice, may also take on the semantic description information that other modes determine image, such as figure 7 below
Shown.
Fig. 7 be a diagram that the second of the semantic description information of the determination image in the image processing method of Fig. 1
The flow chart of example.As shown in fig. 7, the described locus based on described image block in described image
The semantic description information (S130) determining described image with the semantic description information of described image block may include：
Semantic description information based on locus in described image for the described image block and described image block determines
The adjacent image block with similar semantic description information, as Content aggregation region, described similar semantic is retouched
Information of stating is as the semantic description information (S131A) in described Content aggregation region；According to described Content aggregation
The semantic description information in the region and described Content aggregation region locus in described image determine described
The weight (S132A) in Content aggregation region；Weight according to described Content aggregation region and described content are gathered
The semantic description information in collection region determines the semantic description information (S133A) of described image.
In S131A, the adjacent image block in image with similar semantic description information is gathered as content
Collection region.This Content aggregation region generally includes multiple images block, has bigger area.But,
Compared with the case of horn of plenty, this Content aggregation region potentially includes an image block to the content of image.This is interior
Hold aggregation zone and there is similar semantic description information.For example, the image block B (1,1) in the image of Fig. 2,
B (1,2), B (2,1), B (2,2) have similar semantic description information, can be used as Content aggregation region.
Correspondingly, image is made up of multiple Content aggregation regions.
In S132A, calculate each Content aggregation region weight in the picture.Can scheme using with calculating
Weight similar mode as block to calculate the weight in Content aggregation region.For example, it is possible to be gathered according to content
The semantic description information in collection region calculates described Content aggregation region and is adjacent between Content aggregation region
Similarity；It is adjacent the similarity setting institute between Content aggregation region based on described Content aggregation region
State the weight in Content aggregation region.
The semantic description of image in S133A, is determined with the semantic description information based on image block and weight
As info class, weight according to Content aggregation region and semantic description information determine the semantic description of image
Information, and specifically may refer to the description carrying out above in conjunction with S133.
Be can be seen that according to the above description with reference to Fig. 7 and divide each Content aggregation from the image block of image
Region, and according to obtain the semantic description info class of image based on the semantic description information of image block as
Mode, obtains the semantic description information of image based on the semantic description information in Content aggregation region.Work as image
Larger, when content is relatively enriched, this can reduce information processing capacity, thus quickly determining image
Semantic description information.
In the technical scheme according to the image processing method of the embodiment of the present disclosure, by based on each image
The semantic description information of locus in whole image for the block and each image block determines the language of whole image
Adopted description information, thus organically contacted by the associated picture block in whole image, more meets to image
Understand custom.
Fig. 8 is the block diagram schematically illustrating the first image processing apparatus 800 according to the embodiment of the present disclosure.
As shown in figure 8, the first image processing apparatus 800 include one or more processors 810, storage device
820th, input equipment 830, output device 840, communicator 850 and photographic head 860, these assemblies
By bindiny mechanism's (not shown) interconnection of bus system 870 and/or other forms.It should be noted that figure
The assembly of the first image processing apparatus 800 shown in 8 and structure are illustrative, and not restrictive,
As needed, the first image processing apparatus 800 can also have other assemblies and structure.
Processor 810 can be CPU (CPU) or have data-handling capacity and/or refer to
Make the processing unit of the other forms of executive capability, and can control in the first image processing apparatus 800
Other assemblies to execute desired function.
Storage device 820 can include one or more computer programs, and described computer program produces
Product can include various forms of computer-readable recording mediums, for example volatile memory and/or non-volatile
Property memorizer.Described volatile memory for example can include random access memory (RAM) and/or height
Fast buffer storage (cache) etc..Described nonvolatile memory for example can include read only memory
(ROM), hard disk, flash memory etc..Described computer-readable recording medium can store one or many
Individual computer program instructions, processor 810 can run described program instruction, to realize above in conjunction with figure
The image processing method of 1 to Fig. 7 description.Described computer-readable recording medium can also store each
Kind of application program and various data, such as view data and described application program using and/or produce each
Plant data etc..
Input equipment 830 can be the device for input instruction for the user, and can include keyboard, Mus
One or more of mark, mike and touch screen etc..Described instruction e.g. uses following photographic head 860
The instruction of shooting image.Output device 840 can export various information (examples to outside (such as user)
As image or sound), and one or more of display, speaker etc. can be included.Communicator
850 can be by network or other technology and other device (such as personal computer, server, movements
Platform, base station etc.) communication, described network can be the Internet, WLAN, mobile communications network etc.,
Described other technology for example can include Bluetooth communication, infrared communication etc..Photographic head 860 can shoot and treat
The image (such as photo, video etc.) processing, and captured image is stored in storage device 820
In for other assemblies use.
Fig. 9 is the block diagram schematically illustrating the second image processing apparatus 900 according to the embodiment of the present disclosure.
As shown in figure 9, the second image processing apparatus 900 may include division unit 910, the semantic determination of image block
Unit 920 and image, semantic determining unit 930.Pending image averaging is divided by division unit 910
For multiple images block.Image block semanteme determining unit 920 obtains the semantic description information of described image block.
Image, semantic determining unit 930 is based on locus in described image for the described image block and described image
The semantic description information of block determines the semantic description information of described image.
Pending image averaging is divided into multiple images block by division unit 910.Pending image leads to
Often to be represented with the numerical value of each pixel of image.For each pixel, gray value, three primary colories can be used
Component etc. represents.Image typically comprises the picture element matrix arranging according to row and column mode.As being divided into
The example of multiple images block, can divide successively according to the mode that each image block includes 16 × 16 pixels.
The specific example that divides may refer to the diagram of Fig. 2 and the description carrying out with reference to Fig. 2.In fig. 2 each
Image block represents, wherein i is line number in whole image for the image block B and 1≤i≤7, j with B (i, j)
It is row number in whole image for the image block B and 1≤j≤8.In the application, it is possible to use processor and
Memorizer is realizing this division unit 910.
So that the total pixel number of pending image is divided equally just as a example it is illustrated in Fig. 2.Waiting to locate
In the case of the total pixel number of the image of reason is not the integral multiple of image block, it is also the end from picture element matrix
Start, be expert at 16 × 16 pixels for unit and column direction divides, and by last less than 16 × 16 pictures
The image-region of element is also divided into image block.
Pending image averaging is divided into multiple images block, this dividing mode letter by division unit 910
Single.However, other generate images semantic description information image procossing mode in it may be necessary to
Visual signature based on image is split to image and is obtained multiple images region, and this is extremely complex.
Every in the multiple images block that image block semanteme determining unit 920 is divided for division unit 910
Individual image block, all obtains the semantic description information of this image block.The implication of data is exactly semantic, and semanteme is retouched
The information of stating is used to the information of descriptive semantics.View data inherently symbol, is only endowed implication
Data can be used, and at this time data has translated into semantic description information.This semantic description information
E.g. a kind of can be by the class text language performance of machine intuitivism apprehension.For example, the seashore landscape of Fig. 2
The semantic description information of each image block of photo can be each image block be sky, ocean, in sandy beach
The probability distribution of at least one.
Image block semanteme determining unit 920 can obtain the semantic description information of image block as follows：Obtain with
The gauss hybrid models of each corresponding picture material of semantic description information；According to described image block and described
Gauss hybrid models determine the semantic description information of described image block.
Image block semanteme determining unit 920 can obtain and each semantic description information pair from data base in advance
The training image of each picture material answered, and the training image based on each picture material is obtaining this figure
Gauss hybrid models as content.Or, image block semanteme determining unit 920 can be directly from data base
The middle gauss hybrid models obtaining each picture material.
Prestore the training image of each picture material in data base, described image content for example includes
Sky, ocean, sandy beach, meadow, great Shu, high mountain etc..Described image content is the semanteme of various images
Involved content in description information.Each picture material is likely to be of different images, for example, sky
Color under the conditions of different weather and brightness are typically different.Therefore, corresponding with each picture material
Be likely to be of multiple training images, such as 512.For each picture material, can be many based on this
Individual training image is set up a Gauss model to characterize this picture material.Gauss model utilizes normal distribution curve
Accurately quantized image content, by the gray scale in picture material, color etc. be decomposed into some based on normal state
Distribution curve and the model that formed.For each picture material, corresponding Gauss model is averagely weighted
The gauss hybrid models of this picture material can be obtained.The mode of the acquisition gauss hybrid models taken is not
Constitute the restriction to the embodiment of the present disclosure.Generally, the training image of each picture material and every is obtained ahead of time
The gauss hybrid models of individual picture material, and set up data base.
As an example, image block semanteme determining unit 920 can determine respectively according to gauss hybrid models as follows
The semantic description information of individual image block：Similarity according to described image block and described gauss hybrid models Lai
Determine that described image block belongs to the probability Estimation of each picture material described；Institute is belonged to according to described image block
The probability Estimation stating each picture material determines the semantic description information of described image block.
, by the 16 × 16 of the image block B (1,8) in this upper right corner taking the image block in the upper right corner in Fig. 2 as a example
The view data of pixel is mated with the gauss hybrid models of each picture material being obtained respectively.Example
As the view data of image block B (1,8) is mixed with the gauss hybrid models of sky, the Gauss of ocean respectively
Matched moulds type, the gauss hybrid models at sandy beach, the gauss hybrid models on meadow, the gauss hybrid models of big tree,
The gauss hybrid models of high mountain are mated, with determine respectively image block B (1,8) belong to sky, ocean,
Sandy beach, meadow, great Shu, the probability Estimation of high mountain.According to Fig. 2, the image block in the described upper right corner
It is sky, may thereby determine that image block B (1,8) belongs to the maximum probability of sky, and belong to ocean, sand
Beach, meadow, great Shu, the probability of any one may very little even zero in high mountain.
Hereafter, image block semanteme determining unit 920 each picture material according to described image block belongs to
Probability Estimation determine the semantic description information of described image block.For example, image block semanteme determining unit 920
Can represent that each image block belongs to the probability of each picture material with rectangular histogram, can also be with ratio chart
Show that each image block belongs to the probability of each picture material.Specifically will depending on described semantic description information
Ask, described probability Estimation can be carried out with suitable treatments with the semantic description information required for generating.
For each image block in pending image, image block semanteme determining unit 920 all judges this figure
As described in the similarity of block and the described gauss hybrid models of each picture material to determine, each image block belongs to
In the probability Estimation of each picture material described, and determine that based on described probability Estimation its semantic description is believed
Breath, thus obtain the semantic description information of each image block in pending image.Image block is semantic really
The semantic description information of the image block that order unit 920 is obtained may refer to the diagram of Fig. 4 and combines figure
4 associated description carrying out.
Image block semanteme determining unit 920 can be realized using memorizer and processor.When processor fortune
Each operation of image block semanteme determining unit 920 in line storage during the program of storage, can be completed.
Image, semantic determining unit 930 is based on locus in described image for the described image block and image
The semantic description information of the image block that block semanteme determining unit 920 determines is retouched come the semanteme to determine described image
State information.As an example, image, semantic determining unit 930 can be according to the semantic description of described image block
The information and described image block locus in described image determine the weight of described image block；According to institute
The weight of the semantic description information and described image block of stating image block determines the semantic description letter of described image
Breath.By the locus in the picture according to image block, weight is set for this image block, can distinguish ground
The importance of each image block is set, thus more accurately expressing the semantic description information of whole image.Cause
This, be no longer separate between each image block that division unit 910 is divided, and be intended to be based on
Each image block locus in the picture are determining the semantic description information of image.
Image, semantic determining unit 930 can determine the weight of image block as follows：According to described image block
Semantic description information calculate described image block be adjacent the similarity between image block；Based on described figure
The similarity being adjacent between image block as block arranges the weight of described image block.
Semantic description information between two adjacent image blocks is closer between two adjacent image blocks
Similarity is bigger.For example, the image block B (Isosorbide-5-Nitrae) in Fig. 4 and its adjacent image block B (1,3), B (1,5),
All based on sky, ocean, the content at sandy beach occupy small part to the picture material of B (2,4), so should
Similarity between the image block that image block B (1,4) is adjacent is larger.When calculating similarity, examine
Consider each picture material of image block and the ratio occupied by each picture material.Image, semantic determining unit
930 for example can calculate this by the Euclidean distance between the semantic description information of two image blocks of calculating
Similarity between two image blocks.Image block B (2,4) in Fig. 4 and adjacent image block B (2,3),
B (2,5), the similarity height of B (Isosorbide-5-Nitrae), all based on sky, ocean, the content at sandy beach occupy small part；
But image block B (2,4) is low with the similarity of adjacent image block B (3,4) because image block B (3,4) with
Based on ocean, sky, the content at sandy beach occupy small part.Therefore, image block B (2,4) is adjacent
Similarity between image block is adjacent the similarity between image block less than image block B (1,4).
In the case of the similarity height that image block is adjacent between image block, image, semantic determining unit
930 can arrange high weight for described image block；It is adjacent the similarity between image block in image block
In the case of low, it is that described image block arranges low weight.For example, for above-mentioned image block B (Isosorbide-5-Nitrae) and
Weight W (Isosorbide-5-Nitrae) of B (2,4), image block B (Isosorbide-5-Nitrae) determined by image, semantic determining unit 930 is more than figure
Weight W (2,4) as block B (2,4).The like, thus obtaining the weight of each image block B (i, j)
W(i,j).In the application, image, semantic determining unit 930 may also take on other modes and determines each figure
As the weight of block, for example, can be determined according to the Pixel Information of each image block.
Image, semantic determining unit 930 and then the semantic description information according to described image block and described image
The weight of block determines the semantic description information of described image.Multiple images content is potentially included in whole image,
The diverse location that each picture material is located in image.Correspondingly, the semantic description information of image is figure
The distribution of the semantic description information in the diverse location region of picture.
Continue, image, semantic determining unit 930 is retouched according to the semanteme of image block taking the image in Fig. 2 as a example
State the weight of information and image block it is found that the semanteme of each image block in the first row and the second row is retouched
Information of stating is all based on sky, and the ratio very little at ocean and sandy beach, and each image of the first row
Block has very big weight, and the weight of each image block in the second row is less than the image block in the first row as early as possible
Weight, but also there is higher weight, thus the language based on the first row and each image of the second row
Adopted description information and weight can determine that the semantic description information in this region for the image is sky, and ocean and
Minimum ratio is occupied at sandy beach.Similarly, image, semantic determining unit 930 can determine that other of image
Semantic description information in the band of position.
The semantic description information of image determined by image, semantic determining unit 930 may refer to the figure of Fig. 6
Show and related description.In short, the first row of image and the second row can be with the first semantic description information
To describe, the right side four row in the third line of image, fourth line and fifth line can use the second semantic description
Describing, the left side four in the fifth line of image arranges information, the 6th row, the 7th row can be retouched with the 3rd semanteme
State information to describe.First semantic description information indicates that sky occupies maximum ratio, and ocean and sandy beach are divided
Do not occupy small percentage.Second semantic description information indicates that ocean occupies maximum ratio, sky and sandy beach
Occupy small percentage respectively.3rd semantic description information indicates that sandy beach occupies maximum ratio, sky and sea
Ocean occupies small percentage respectively.
Alternatively, image, semantic determining unit 930 may also take on the semanteme that other modes determine image
Description information.For example, image, semantic determining unit 930 can be based on described image block in described image
Locus and described image block semantic description information determine there is the adjacent of similar semantic description information
Image block as Content aggregation region, using described similar semantic description information as described Content aggregation region
Semantic description information；Semantic description information according to described Content aggregation region and described Content aggregation area
Locus in described image for the domain determine the weight in described Content aggregation region；Gathered according to described content
The semantic description information in the weight in collection region and described Content aggregation region determines the semantic description of described image
Information.
Here, image, semantic determining unit 930 will have the neighbor map of similar semantic description information in image
As block is as Content aggregation region.This Content aggregation region generally includes multiple images block, has bigger
Area.But, in the case of the content of image is compared with horn of plenty, this Content aggregation region potentially includes one
Individual image block.Then, image, semantic determining unit 930 is using the side similar with the weight calculating image block
Formula is calculating the weight in Content aggregation region, and the weight according to Content aggregation region and semantic description information
Determine the semantic description information of image.For example, image, semantic determining unit 930 can be according to Content aggregation
The semantic description information in region calculates described Content aggregation region and is adjacent the phase between Content aggregation region
Like degree；The similarity setting being adjacent between Content aggregation region based on described Content aggregation region is described
The weight in Content aggregation region, the weight according to Content aggregation region and semantic description information determine image
Semantic description information.That is, whole image is drawn by image, semantic determining unit 930 based on image block
Divide into Content aggregation region, and according to the semanteme obtaining image based on the semantic description information of image block
Description information similar mode, is retouched based on the semanteme that the semantic description information in Content aggregation region obtains image
State information.When image is larger, when content is relatively enriched, this can reduce information processing capacity, thus
Quickly determine the semantic description information of image.
Image, semantic determining unit 930 can be realized using memorizer and processor.When processor runs
Each operation of image, semantic determining unit 930 in memorizer during the program of storage, can be completed.
In the technical scheme according to the image processing apparatus of the embodiment of the present disclosure, by based on each image
The semantic description information of locus in whole image for the block and each image block determines the language of whole image
Adopted description information, thus organically contacted by the associated picture block in whole image, more meets to image
Understand custom.
Those skilled in the art can be understood that, for convenience and simplicity of description, above-mentioned retouches
The device stated, the specific work process of unit, may be referred to the corresponding process in preceding method embodiment,
Will not be described here.
Those of ordinary skill in the art are it is to be appreciated that combine each of the embodiments described herein description
The unit of example and step, can be come with the combination of electronic hardware or computer software and electronic hardware
Realize.These functions to be executed with hardware or software mode actually, and specific depending on technical scheme should
With and design constraint.Professional and technical personnel can use different methods to each specific application
Realize described function, but this realization is it is not considered that exceed the scope of the present disclosure.
It should be understood that disclosed method and apparatus in several embodiments provided herein,
Can realize by another way.For example, apparatus embodiments described above are only schematically,
For example, the division of described unit, only a kind of division of logic function, in addition actual can have when realizing
Dividing mode, for example multiple units can in conjunction with or be desirably integrated into another system, or some are special
Levy and can ignore, or do not execute.
The above, the only specific embodiment of the disclosure, but the protection domain of the disclosure does not limit to
In this, any those familiar with the art, can be easily in the technical scope that the disclosure discloses
Expect change or replacement, all should cover within the protection domain of the disclosure.Therefore, the protection of the disclosure
Scope should described be defined by scope of the claims.
Claims (12)
1. a kind of image processing method, including：
Pending image averaging is divided into multiple images block；
Obtain the semantic description information of described image block；
Semantic description information based on locus in described image for the described image block and described image block
Determine the semantic description information of described image.
2. image processing method according to claim 1, wherein, described acquisition described image block
Semantic description information includes：
Obtain the gauss hybrid models with each corresponding picture material of semantic description information；
Determine the semantic description information of described image block according to described image block and described gauss hybrid models.
3. image processing method according to claim 2, wherein, described according to described image block and
Described gauss hybrid models determine that the semantic description information of described image block includes：
Determine that described image block belongs to institute according to the similarity of described image block and described gauss hybrid models
State the probability Estimation of each picture material；
According to described image block belongs to, the probability Estimation of each picture material determines the language of described image block
Adopted description information.
4. image processing method according to claim 1, wherein, described is existed based on described image block
The semantic description information of the locus in described image and described image block determines that the semanteme of described image is retouched
The information of stating includes：
Locus in described image for the semantic description information and described image block according to described image block
Determine the weight of described image block；
The weight of the semantic description information according to described image block and described image block determines the language of described image
Adopted description information.
5. image processing method according to claim 4, wherein, described according to described image block
The semantic description information and described image block locus in described image determine the weight of described image block
Including：
Semantic description information according to described image block calculates described image block and is adjacent between image block
Similarity；
The similarity being adjacent between image block based on described image block arranges the weight of described image block.
6. image processing method according to claim 1, wherein, described is existed based on described image block
The semantic description information of the locus in described image and described image block determines that the semanteme of described image is retouched
The information of stating includes：
Semantic description information based on locus in described image for the described image block and described image block
Determine the adjacent image block with similar semantic description information as Content aggregation region, by described similar language
Adopted description information is as the semantic description information in described Content aggregation region；
Semantic description information according to described Content aggregation region and described Content aggregation region are in described image
In locus determine the weight in described Content aggregation region；
The semantic description information in the weight according to described Content aggregation region and described Content aggregation region determines
The semantic description information of described image.
7. a kind of image processing apparatus, including：
Memorizer；With
Processor, for executing following operation：
Pending image averaging is divided into multiple images block；
Obtain the semantic description information of described image block；
Semantic description based on locus in described image for the described image block and described image block
Information determines the semantic description information of described image.
8. image processing apparatus according to claim 7, wherein, described acquisition described image block
Semantic description information includes：
Obtain the gauss hybrid models with each corresponding picture material of semantic description information；
Determine the semantic description information of described image block according to described image block and described gauss hybrid models.
9. image processing apparatus according to claim 8, wherein, described according to described image block and
Described gauss hybrid models determine that the semantic description information of described image block includes：
Determine that described image block belongs to institute according to the similarity of described image block and described gauss hybrid models
State the probability Estimation of each picture material；
According to described image block belongs to, the probability Estimation of each picture material determines the language of described image block
Adopted description information.
10. image processing apparatus according to claim 7, wherein, described based on described image block
The semantic description information of the locus in described image and described image block determines the semanteme of described image
Description information includes：
Locus in described image for the semantic description information and described image block according to described image block
Determine the weight of described image block；
The weight of the semantic description information according to described image block and described image block determines the language of described image
Adopted description information.
11. image processing apparatus according to claim 10, wherein, described according to described image block
Locus in described image of semantic description information and described image block determine the power of described image block
Include again：
Semantic description information according to described image block calculates described image block and is adjacent between image block
Similarity；
The similarity being adjacent between image block based on described image block arranges the weight of described image block.
12. image processing apparatus according to claim 7, wherein, described based on described image block
The semantic description information of the locus in described image and described image block determines the semanteme of described image
Description information includes：
Semantic description information based on locus in described image for the described image block and described image block
Determine the adjacent image block with similar semantic description information as Content aggregation region, by described similar language
Adopted description information is as the semantic description information in described Content aggregation region；
Semantic description information according to described Content aggregation region and described Content aggregation region are in described image
In locus determine the weight in described Content aggregation region；
The semantic description information in the weight according to described Content aggregation region and described Content aggregation region determines
The semantic description information of described image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510508531.6A CN106469437B (en) | 2015-08-18 | 2015-08-18 | Image processing method and image processing apparatus |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510508531.6A CN106469437B (en) | 2015-08-18 | 2015-08-18 | Image processing method and image processing apparatus |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106469437A true CN106469437A (en) | 2017-03-01 |
| CN106469437B CN106469437B (en) | 2020-08-25 |
Family
ID=58214749
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510508531.6A Active CN106469437B (en) | 2015-08-18 | 2015-08-18 | Image processing method and image processing apparatus |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106469437B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107133951A (en) * | 2017-05-22 | 2017-09-05 | 中国科学院自动化研究所 | Distorted image detection method and device |
| CN107273863A (en) * | 2017-06-21 | 2017-10-20 | 天津师范大学 | A kind of scene character recognition method based on semantic stroke pond |
| CN113889074A (en) * | 2021-09-27 | 2022-01-04 | 平安科技（深圳）有限公司 | Voice generation method, device, equipment and medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102222239A (en) * | 2011-06-03 | 2011-10-19 | 哈尔滨工程大学 | Labelling image scene clustering method based on vision and labelling character related information |
| US20120141019A1 (en) * | 2010-12-07 | 2012-06-07 | Sony Corporation | Region description and modeling for image subscene recognition |
-
2015
- 2015-08-18 CN CN201510508531.6A patent/CN106469437B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120141019A1 (en) * | 2010-12-07 | 2012-06-07 | Sony Corporation | Region description and modeling for image subscene recognition |
| CN102222239A (en) * | 2011-06-03 | 2011-10-19 | 哈尔滨工程大学 | Labelling image scene clustering method based on vision and labelling character related information |
Non-Patent Citations (3)
| Title |
|---|
| 付杰: "基于场景语义图像标注关键技术的研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 刘硕研 等: "一种基于上下文语义信息的图像块视觉单词生成算法", 《电子学报》 * |
| 陈娜: "基于高斯混合模型的自动图像标注方法", 《计算机应用》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107133951A (en) * | 2017-05-22 | 2017-09-05 | 中国科学院自动化研究所 | Distorted image detection method and device |
| CN107133951B (en) * | 2017-05-22 | 2020-02-28 | 中国科学院自动化研究所 | Image tampering detection method and device |
| CN107273863A (en) * | 2017-06-21 | 2017-10-20 | 天津师范大学 | A kind of scene character recognition method based on semantic stroke pond |
| CN107273863B (en) * | 2017-06-21 | 2019-07-23 | 天津师范大学 | A kind of scene character recognition method based on semantic stroke pond |
| CN113889074A (en) * | 2021-09-27 | 2022-01-04 | 平安科技（深圳）有限公司 | Voice generation method, device, equipment and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106469437B (en) | 2020-08-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109618222B (en) | A kind of splicing video generation method, device, terminal device and storage medium | |
| CN109919869B (en) | Image enhancement method and device and storage medium | |
| US10134165B2 (en) | Image distractor detection and processing | |
| Chen et al. | Cross parallax attention network for stereo image super-resolution | |
| CN112651438A (en) | Multi-class image classification method and device, terminal equipment and storage medium | |
| CN110516096A (en) | Synthesis perception digital picture search | |
| CN111062854B (en) | Method, device, terminal and storage medium for detecting watermark | |
| WO2022028313A1 (en) | Method and device for image generation and colorization | |
| Zhang et al. | A comprehensive survey on computational aesthetic evaluation of visual art images: Metrics and challenges | |
| CN111932431B (en) | Visible watermark removing method based on watermark decomposition model and electronic equipment | |
| Joshi | OpenCV with Python by example | |
| WO2023056835A1 (en) | Video cover generation method and apparatus, and electronic device and readable medium | |
| CN111127309A (en) | Portrait style transfer model training method, portrait style transfer method and device | |
| CN113822830A (en) | Multi-exposure image fusion method based on depth perception enhancement | |
| Zhang et al. | Retargeting semantically-rich photos | |
| KR20160074958A (en) | Apparatus and method for generating motion effects by analyzing motion of object | |
| CN115294055A (en) | Image processing method, image processing device, electronic equipment and readable storage medium | |
| CN106469437A (en) | Image processing method and image processing apparatus | |
| CN117252791A (en) | Image processing method, device, electronic equipment and storage medium | |
| Song et al. | Photo squarization by deep multi-operator retargeting | |
| CN113506305A (en) | Image enhancement method, semantic segmentation method and device for three-dimensional point cloud data | |
| Stahl et al. | Ist-style transfer with instance segmentation | |
| Togo et al. | Text-guided style transfer-based image manipulation using multimodal generative models | |
| Jiang et al. | Deep fusion local-content and global-semantic for image inpainting | |
| Xiang et al. | Panoramic image style transfer technology based on multi-attention fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |