CN104636759B - A kind of method and picture filter information recommendation system for obtaining picture and recommending filter information - Google Patents
A kind of method and picture filter information recommendation system for obtaining picture and recommending filter information Download PDFInfo
- Publication number
- CN104636759B CN104636759B CN201510090429.9A CN201510090429A CN104636759B CN 104636759 B CN104636759 B CN 104636759B CN 201510090429 A CN201510090429 A CN 201510090429A CN 104636759 B CN104636759 B CN 104636759B
- Authority
- CN
- China
- Prior art keywords
- filter
- pattern recognition
- recognition model
- model
- filter information
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Abstract
The invention discloses methods and picture filter information recommendation system that a kind of acquisition picture recommends filter information, it is related to image processing technology, it is intended to provide a kind of method and system that smart filter is recommended, technical key point includes that filter information is split as to multiple filter mirror properties, largely to add the picture of filter to go to train multiple pattern recognition models as training sample, each model is made to have the ability for identifying according to picture feature and being suitble to the value for filtering mirror properties of the picture.The value for the filter mirror properties that each model exports finally is integrated into the filter information recommended.
Description
Technical field
The present invention relates to image processing technologies.
Background technique
Most of at present to take pictures application program and all provide the function of increasing filter for picture for user, user can be with
Increase filter according to the picture that the hobby of oneself is shooting to achieve the purpose that beautify picture.As the mobile terminal APP(that takes pictures is answered
With program, foreign language full name: Application) continuous development, in order to provide the user with more wide creation space, much
APP is that the filter edition interface that provides of user is more and more, for example, have special efficacy, aperture, texture, color, colour cast, level, skin makeup,
Clarity, colour temperature, dark angle etc..It is most of but when the so how editable sub- property parameters of filter are put in front of the user
User cannot but utilize these functions very well, and it is allowed to think that this is a very complicated, professional very strong tool.
In order to solve this problem, occur many filter proposed algorithms in the prior art, main thought is according to figure
Piece scene content directly recommends possible filter to user, but the result recommended in this way is very general, and user experience is simultaneously bad.
Because when it come to arriving the scope of artistry, using single causality (such as scene content and filter) come when describing, always
A large amount of important informations can be omitted.
Summary of the invention
Based on such realistic problem, the present invention is intended to provide the method and system that a kind of smart filter is recommended, basic to think
Want for filter information to be split as multiple filter mirror properties that (actually each filter is described by one group of filter mirror properties
), with largely added filter picture (including original image information, i.e., the picture or photo handled without later image, and
The filter information that user adds for it) it goes to train multiple pattern recognition models (each pattern recognition model use as training sample
In one filter mirror properties of identification), have each model and identifies the filter mirror properties for being suitble to the picture according to picture feature
The ability of value.The value for the filter mirror properties that each model exports finally is integrated into the filter information recommended.
It is realized substantially based on foregoing invention, the particular technique means that the present invention uses include:
Step 1: obtaining a certain number of samples pictures, extract original image information and filter letter in every samples pictures
Breath;The filter information includes several filter mirror properties;
Step 2: according to the feature vector of every samples pictures of original image information extraction in every samples pictures;
Step 3: selecting pattern recognition model to be trained;The quantity of the pattern recognition model and the filter information
The sub- number of attributes of filter is identical, and the corresponding filter mirror properties of each pattern recognition model;
Step 4: training each pattern recognition model in accordance with the following methods: successively with the feature vector of every samples pictures
For the input of pattern recognition model, corresponding filter mirror properties the taking in the samples pictures filter information of identification model in mode
Value is the output training of the pattern recognition model pattern recognition model;
Step 5: obtaining the target original image of filter information to be recommended;
Step 6: extracting the feature vector of target original image;
Step 7: the feature vector of target original image is sequentially input into the pattern recognition model after each training, Ge Gexun
Pattern recognition model after white silk exports the value of its corresponding filter mirror properties;Filter that each pattern recognition model is exported
The value of attribute is integrated the filter information recommended.
Preferably, described eigenvector includes color space characteristic value, texture eigenvalue and Structural Eigenvalue.
Preferably, the filter mirror properties include texture, it is aperture, special efficacy, colour temperature, tone, exposure, contrast, bright-coloured
Degree and bloom.
Preferably, the pattern recognition model includes classification mode identification model and Regression Model identification model;Corresponding line
The pattern recognition model of reason, aperture and special efficacy three filter mirror properties is classification mode identification model;Corresponding colour temperature, tone, exposure
The pattern recognition model that luminosity, contrast, vividness and bloom six filter mirror properties is Regression Model identification model.
The present invention also provides a kind of picture filter information recommendation systems, comprising:
Target original image acquiring unit, for obtaining the target original image of filter information to be recommended;
Target original image characteristic vector pickup unit, for extracting the feature vector of target original image;
The sub- Attribute Recognition unit of filter, for the feature vector of target original image to be sequentially input each pattern-recognition mould
Type obtains the value of the filter mirror properties of each pattern recognition model output;Wherein, each pattern recognition model is one corresponding
Mirror properties are filtered, for calculating the value of its corresponding filter mirror properties according to the feature vector of target original image;
The value of filter information integral unit, the filter mirror properties for exporting each pattern recognition model is integrated
The filter information recommended.
Further, each pattern recognition model obtains in the following way:
Step 1: obtaining a certain number of samples pictures, extract original image information and filter letter in every samples pictures
Breath;The filter information includes several filter mirror properties;
Step 2: according to the feature vector of every samples pictures of original image information extraction in every samples pictures;
Step 3: selecting pattern recognition model to be trained;The quantity of the pattern recognition model and the filter information
The sub- number of attributes of filter is identical, and the corresponding filter mirror properties of each pattern recognition model;
Step 4: training each pattern recognition model in accordance with the following methods: successively with the feature vector of every samples pictures
For the input of pattern recognition model, corresponding filter mirror properties the taking in the samples pictures filter information of identification model in mode
Value is the output training of the pattern recognition model pattern recognition model.
Further, further includes:
Samples pictures information extraction unit is extracted in every samples pictures for obtaining a certain number of samples pictures
Original image information and filter information;The filter information includes several filter mirror properties;
Samples pictures characteristic vector pickup unit, for according to the original image information extraction every in every samples pictures
The feature vector of samples pictures;
Pattern recognition model determination unit, for selecting pattern recognition model to be trained;The pattern recognition model
Quantity is identical as the sub- number of attributes of filter of the filter information, and the corresponding filter category of each pattern recognition model
Property;
Pattern recognition model training unit, for training each pattern recognition model in accordance with the following methods: successively with every
The feature vector for opening samples pictures is the input of pattern recognition model, and the corresponding filter mirror properties of identification model are in the sample in mode
Value in this picture filter information is the output training of the pattern recognition model pattern recognition model.
Due to using above-mentioned technological means, the invention has the following advantages:
1. filter information is decomposed into several filter mirror properties for describing it by the present invention, for each filter mirror properties training one
A pattern recognition model, the original image new for every, the present invention in pattern recognition model according to its feature vector calculate
Recommend a filter mirror properties for it, then each filter mirror properties are integrated into filter information, the filter recommended in this way for picture
Information is more accurate, proper, and user experience is good.
2. the feature vector of picture includes color space characteristic value, texture eigenvalue and Structural Eigenvalue in the present invention, more
Add and comprehensively describe the feature of original image, ensure that the accuracy of proposed algorithm, at the same these three characteristic value calculation amounts compared with
It is small, reduce resource cost, the present invention is made not only to have can be suitably used for computer but also can be suitably used for all kinds of intelligent movable equipment.
3. the present invention has selected different pattern recognition types for the numeric form (discrete or continuous) of filter mirror properties,
Ensure the recognition accuracy of filter mirror properties.
Detailed description of the invention
Examples of the present invention will be described by way of reference to the accompanying drawings, in which:
Fig. 1 is the flow chart of one specific embodiment of the method for the present invention.
Specific embodiment
All features disclosed in this specification or disclosed all methods or in the process the step of, in addition to mutually exclusive
Feature and/or step other than, can combine in any way.
Any feature disclosed in this specification unless specifically stated can be equivalent or with similar purpose by other
Alternative features are replaced.That is, unless specifically stated, each feature is an example in a series of equivalent or similar characteristics
?.
Such as Fig. 1, a specific embodiment of the invention includes the foundation of the sub- attribute Recognition Model of filter and is that target is original
Picture recommends two stages of filter information.
The process for wherein establishing the sub- attribute Recognition Model of filter includes:
Step 1: obtaining a certain number of samples pictures, and extract original image information and filter in every samples pictures
Information.
Here required samples pictures refer to that user thinks that it increased the picture of filtering effects, wrap in description information
Information containing original image and filter information.Wherein, original image information refers to processed without subsequent image processing technique
The information of picture, the information such as R, G, B value for example including each pixel of original image, filter information include the filter for describing it
Sub- attribute.The different APP that take pictures are that the open editable filter mirror properties of user are different, such as filter mirror properties include but not
It is limited to texture (Lighting), aperture (TiltShift), special efficacy (filter), level (EnhanceHdr), skin makeup
(EnhanceSkin), clarity (sharpness), colour temperature (Temperature), tone (Hue), exposure
(Exposure), contrast (Contrast), vividness (Vibrance), saturation degree (Saturation), bloom
(HighLight), shade (Shadow), dark angle (VignetteStrong), center brightness (CenterStrong).
Certain amount described here can be several hundred, be also possible to a thousand sheets.The quantity of samples pictures is bigger, source
Abundanter, the pattern recognition model obtained using the training of these samples pictures is more accurate.
Step 2: according to the feature vector of every samples pictures of original image information extraction in every samples pictures.
The extraction of picture feature is a more complex field, it is contemplated that real-time and calculation amount, present invention selection
Calculation amount is relatively small, but has the feature with distinction, including color space feature, and textural characteristics and structure feature three are big
Category feature.The extraction of this three categories feature is all made of the prior art, below 2 say the methods that the present embodiment uses, but should not
Select excellent method as the limitation of picture feature vector in the present invention the present embodiment.
Color space feature
The present embodiment has selected RGB, the bin histogram and mean value and variance of tri- color spaces of HSV, LAB.Together
When also add the bin histogram of grayscale image.As soon as these values are formed a subvector, such color space has 1586
A characteristic value: gray_hist (32)+RGB_hist (512)+HSV_hist (512)+LAB_hist (512)+RGB_
Mean_std (6)+HSV_mean_std (6)+LAB_mean_std (6), wherein the bin histogram gray_ of grayscale image
Hist has 32 characteristic values;The bin histogram RGB_hist of RGB color, hsv color space bin histogram
The bin histogram LAB_hist of HSV_hist, LAB color space all has 512 characteristic values;The mean value of RGB color,
Variance RGB_mean_std, the mean value in hsv color space, the mean value of variance HSV_mean_std, LAB color space, variance
LAB_mean_std all has 6 characteristic values.
In other embodiments, the characteristic value of color space can only include a part of above-mentioned several characteristic values, can also
With bin histogram, mean value and/or the variance etc. for increasing other color spaces on the basis of above-mentioned several characteristic values.
Textural characteristics
The present embodiment handles picture to obtain 32 features using gabor kernel method.Specific kernel method is Mr.
At 16 cores, convolution is sought using each core and picture, in the whole mean value and variance for calculating convolution results, to obtain 32 spies
Value indicative.
In other embodiments, the texture eigenvalue that 8 gabor cores extract picture can be used.
Structure feature
The present embodiment uses HOG algorithm (histograms of oriented gradients, Histogram of Oriented Gradient), mentions
128 features of a picture are taken.The specific practice is to have selected 8 directions, such as upper and lower, left and right, 45 ° of upper left corner side
To, 45 ° of lower left corner direction, 45 ° of upper right corner direction and 45 ° of lower right corner direction.Picture is divided into several cell, each cell
Include 32 × 32 pixels;Each block includes 1 × 1 cell.
In other embodiments, 4 directions can be only selected in HOG algorithm.
Step 3: selecting pattern recognition model to be trained;The quantity of the pattern recognition model and the filter information
The sub- number of attributes of filter is identical, and the corresponding filter mirror properties of each pattern recognition model.
Pattern recognition model has many types, the pattern-recognition that can be used for classifying including for output valve being discrete value
Model, such as SVM(supporting vector machine model) and output valve be successive value Regression Model identification model, such as multiple linear regression mould
Type.
The present embodiment is according to the characteristics taking value of filter mirror properties, and being the different sub- Attributions selections of filter, different modes is known
Other model.If tri- filter mirror properties values of Lighting, TiltShift and ilter are discrete value, the present embodiment is selected thus
Output valve be discrete value can be used for classify pattern recognition model, Temperature, Hue, Exposure, Contrast,
The value that Vibrance and Highlight etc. filters mirror properties is successive value, and it is successive value that the present embodiment, which selects output valve, thus
Regression Model identification model.
Step 4: training each pattern recognition model in accordance with the following methods: successively with the feature vector of every samples pictures
For the input of pattern recognition model, corresponding filter mirror properties the taking in the samples pictures filter information of identification model in mode
Value is the output training of the pattern recognition model pattern recognition model.When samples pictures quantity is more, train
Model recognition accuracy is higher.
Step 5: by taking pictures, the modes such as screenshot obtain the target original image of filter information to be recommended.
Step 6: the feature vector of target original image is extracted according to the same method of step 2.
Step 7: the feature vector of target original image is sequentially input into the pattern recognition model after each training, Ge Gexun
Pattern recognition model after white silk will automatically calculate and export the value of its corresponding filter mirror properties;By each pattern recognition model
The value of the filter mirror properties of output is integrated the filter information recommended.
The present invention also provides a kind of picture filter information recommendation system, which can be directly mounted at intelligent movable equipment
On, comprising:
Target original image acquiring unit, for obtaining the target original image of filter information to be recommended.
Target original image characteristic vector pickup unit, for extracting the feature vector of target original image.
The sub- Attribute Recognition unit of filter, for the feature vector of target original image to be sequentially input each pattern-recognition mould
Type obtains the value of the filter mirror properties of each pattern recognition model output;Wherein, each pattern recognition model is trained in advance
It is ripe, the corresponding filter mirror properties of each pattern recognition model, for calculating it according to the feature vector of target original image
The value of corresponding filter mirror properties.
The value of filter information integral unit, the filter mirror properties for exporting each pattern recognition model is integrated
The filter information recommended.
Wherein, each pattern recognition model obtains in the following way:
Step 1: obtaining a certain number of samples pictures, extract original image information and filter letter in every samples pictures
Breath;The filter information includes several filter mirror properties;
Step 2: according to the feature vector of every samples pictures of original image information extraction in every samples pictures;
Step 3: selecting pattern recognition model to be trained;The quantity of the pattern recognition model and the filter information
The sub- number of attributes of filter is identical, and the corresponding filter mirror properties of each pattern recognition model;
Step 4: training each pattern recognition model in accordance with the following methods: successively with the feature vector of every samples pictures
For the input of pattern recognition model, corresponding filter mirror properties the taking in the samples pictures filter information of identification model in mode
Value is the output training of the pattern recognition model pattern recognition model.
The invention is not limited to specific embodiments above-mentioned.The present invention, which expands to, any in the present specification to be disclosed
New feature or any new combination, and disclose any new method or process the step of or any new combination.
Claims (8)
1. a kind of method for obtaining picture and recommending filter information characterized by comprising
Step 1: obtaining a certain number of samples pictures, extract original image information and filter information in every samples pictures;
The filter information includes the sub- attribute of filter;
Step 2: according to the feature vector of every samples pictures of original image information extraction in every samples pictures;
Step 3: selecting pattern recognition model to be trained;The filter of the quantity of the pattern recognition model and the filter information
Sub- number of attributes is identical, and the corresponding filter mirror properties of each pattern recognition model;
Step 4: each pattern recognition model of training, training method are as follows: successively using the feature vector of every samples pictures as mould
The input of formula identification model, in mode corresponding value of the filter mirror properties in the samples pictures filter information of identification model be
The output training of the pattern recognition model pattern recognition model;
Step 5: obtaining the target original image of filter information to be recommended;
Step 6: extracting the feature vector of target original image;
Step 7: the feature vector of target original image being sequentially input into the pattern recognition model after each training, after each training
Pattern recognition model export its it is corresponding filter mirror properties value;The filter mirror properties that each pattern recognition model is exported
Value integrated the filter information recommended.
2. it is according to claim 1 it is a kind of obtain picture recommend filter information method, which is characterized in that the feature to
Amount includes color space characteristic value, texture eigenvalue and Structural Eigenvalue.
3. a kind of method for obtaining picture and recommending filter information according to claim 1, which is characterized in that filter
Attribute includes texture, aperture, special efficacy, colour temperature, tone, exposure, contrast, vividness and bloom.
4. a kind of method for obtaining picture and recommending filter information according to claim 3, which is characterized in that the mode is known
Other model includes classification mode identification model and Regression Model identification model;Corresponding three texture, aperture and special efficacy filter categories
Property pattern recognition model be classification mode identification model;Corresponding colour temperature, tone, exposure, contrast, vividness and bloom six
The pattern recognition model of a filter mirror properties is Regression Model identification model.
5. a kind of picture filter information recommendation system characterized by comprising
Samples pictures information extraction unit is extracted original in every samples pictures for obtaining a certain number of samples pictures
Pictorial information and filter information;The filter information includes the sub- attribute of filter;
Samples pictures characteristic vector pickup unit, for according to every sample of original image information extraction in every samples pictures
The feature vector of picture;
Pattern recognition model determination unit, for selecting pattern recognition model to be trained;The quantity of the pattern recognition model
It is identical as the sub- number of attributes of filter of the filter information, and the corresponding filter mirror properties of each pattern recognition model;
Pattern recognition model training unit, for training each pattern recognition model, training method are as follows: successively with every sample
The feature vector of picture is the input of pattern recognition model, and the corresponding filter mirror properties of identification model are in the samples pictures in mode
Value in filter information is the output training of the pattern recognition model pattern recognition model;
Target original image acquiring unit, for obtaining the target original image of filter information to be recommended;
Target original image characteristic vector pickup unit, for extracting the feature vector of target original image;
The sub- Attribute Recognition unit of filter, for the feature vector of target original image to be sequentially input each pattern recognition model,
Obtain the value of the filter mirror properties of each pattern recognition model output;Wherein, the corresponding filter of each pattern recognition model
Mirror properties, for calculating the value of its corresponding filter mirror properties according to the feature vector of target original image;
The value of filter information integral unit, the filter mirror properties for exporting each pattern recognition model is integrated to obtain
The filter information of recommendation.
6. a kind of picture filter information recommendation system according to claim 5, which is characterized in that described eigenvector includes
Color space characteristic value, texture eigenvalue and Structural Eigenvalue.
7. a kind of picture filter information recommendation system according to claim 5, which is characterized in that the filter mirror properties packet
Include texture, aperture, special efficacy, colour temperature, tone, exposure, contrast, vividness and bloom.
8. a kind of picture filter information recommendation system according to claim 7, which is characterized in that the pattern recognition model
Including classification mode identification model and Regression Model identification model;The mode of corresponding texture, aperture, special efficacy three filter mirror properties
Identification model is classification mode identification model;Corresponding colour temperature, tone, exposure, six contrast, vividness and bloom filter
The pattern recognition model of attribute is Regression Model identification model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510090429.9A CN104636759B (en) | 2015-02-28 | 2015-02-28 | A kind of method and picture filter information recommendation system for obtaining picture and recommending filter information |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510090429.9A CN104636759B (en) | 2015-02-28 | 2015-02-28 | A kind of method and picture filter information recommendation system for obtaining picture and recommending filter information |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104636759A CN104636759A (en) | 2015-05-20 |
CN104636759B true CN104636759B (en) | 2019-01-15 |
Family
ID=53215487
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510090429.9A Active CN104636759B (en) | 2015-02-28 | 2015-02-28 | A kind of method and picture filter information recommendation system for obtaining picture and recommending filter information |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104636759B (en) |
Families Citing this family (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105224950A (en) * | 2015-09-29 | 2016-01-06 | 小米科技有限责任公司 | The recognition methods of filter classification and device |
CN105303517B (en) * | 2015-10-26 | 2018-09-11 | 北京金山安全软件有限公司 | Image processing method and device |
CN105653142A (en) * | 2016-01-07 | 2016-06-08 | 厦门美图之家科技有限公司 | Image special effect ordering method and system based on areas and user attributes |
CN106778722A (en) * | 2016-11-25 | 2017-05-31 | 天方创新(北京)信息技术有限公司 | Image processing method and device |
CN106778721A (en) * | 2016-11-25 | 2017-05-31 | 天方创新(北京)信息技术有限公司 | Image processing method and device |
CN107517348A (en) * | 2017-08-30 | 2017-12-26 | 广东欧珀移动通信有限公司 | The rendering intent and device of image |
CN107730461A (en) * | 2017-09-29 | 2018-02-23 | 北京金山安全软件有限公司 | Image processing method, apparatus, device and medium |
CN108079579B (en) * | 2017-12-28 | 2021-09-28 | 珠海豹好玩科技有限公司 | Image processing method and device and terminal |
US11093839B2 (en) * | 2018-04-13 | 2021-08-17 | Fujifilm Business Innovation Corp. | Media object grouping and classification for predictive enhancement |
CN109447958B (en) * | 2018-10-17 | 2023-04-14 | 腾讯科技(深圳)有限公司 | Image processing method, image processing device, storage medium and computer equipment |
CN109727208A (en) * | 2018-12-10 | 2019-05-07 | 北京达佳互联信息技术有限公司 | Filter recommended method, device, electronic equipment and storage medium |
CN109741283A (en) * | 2019-01-23 | 2019-05-10 | 芜湖明凯医疗器械科技有限公司 | A kind of method and apparatus for realizing smart filter |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050020903A1 (en) * | 2003-06-25 | 2005-01-27 | Sriram Krishnan | Systems and methods for automated diagnosis and decision support for heart related diseases and conditions |
CN102930454A (en) * | 2012-10-07 | 2013-02-13 | 乐配(天津)科技有限公司 | Intelligent 3D (Three Dimensional) advertisement recommendation method based on multiple perception technologies |
CN103761295A (en) * | 2014-01-16 | 2014-04-30 | 北京雅昌文化发展有限公司 | Automatic picture classification based customized feature extraction algorithm for art pictures |
CN104274164A (en) * | 2013-07-05 | 2015-01-14 | 广州华久信息科技有限公司 | Blood pressure predicting method and mobile phone based on facial image |
-
2015
- 2015-02-28 CN CN201510090429.9A patent/CN104636759B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050020903A1 (en) * | 2003-06-25 | 2005-01-27 | Sriram Krishnan | Systems and methods for automated diagnosis and decision support for heart related diseases and conditions |
CN102930454A (en) * | 2012-10-07 | 2013-02-13 | 乐配(天津)科技有限公司 | Intelligent 3D (Three Dimensional) advertisement recommendation method based on multiple perception technologies |
CN104274164A (en) * | 2013-07-05 | 2015-01-14 | 广州华久信息科技有限公司 | Blood pressure predicting method and mobile phone based on facial image |
CN103761295A (en) * | 2014-01-16 | 2014-04-30 | 北京雅昌文化发展有限公司 | Automatic picture classification based customized feature extraction algorithm for art pictures |
Also Published As
Publication number | Publication date |
---|---|
CN104636759A (en) | 2015-05-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104636759B (en) | A kind of method and picture filter information recommendation system for obtaining picture and recommending filter information | |
Mavridaki et al. | A comprehensive aesthetic quality assessment method for natural images using basic rules of photography | |
CN108206917B (en) | Image processing method and device, storage medium and electronic device | |
WO2019100282A1 (en) | Face skin color recognition method, device and intelligent terminal | |
CN101231757A (en) | Apparatus and method for analyzing picture dominant hue as well as application in picture searching thereof | |
CN103971134B (en) | Image classification, retrieval and bearing calibration, and related device | |
CN108399052A (en) | Picture compression method, apparatus, computer equipment and storage medium | |
Čuljak et al. | Classification of art paintings by genre | |
CN103914708A (en) | Food variety detection method and system based on machine vision | |
US8559714B2 (en) | Post processing for improved generation of intrinsic images | |
CN108647696B (en) | Picture color value determining method and device, electronic equipment and storage medium | |
US8498496B2 (en) | Method and apparatus for filtering red and/or golden eye artifacts | |
CN101256597B (en) | Method and apparatus for recognizing color blankness reduced graph as well as application thereof in search index | |
CN107886549B (en) | Fabric pattern color transfer method based on weaving texture detail enhancement | |
CN108416782A (en) | View-based access control model identifies and the tobacco leaf ranking method and system of illumination correction | |
US8971619B2 (en) | Method and a device for extracting color features | |
CN110415816B (en) | Skin disease clinical image multi-classification method based on transfer learning | |
US8428352B1 (en) | Post processing for improved generation of intrinsic images | |
US20220358752A1 (en) | Apparatus and method for developing space analysis model based on data augmentation | |
KR100488014B1 (en) | YCrCb color based human face location detection method | |
US8553979B2 (en) | Post processing for improved generation of intrinsic images | |
US11682143B2 (en) | System and method for hair analysis of user | |
CN112218006B (en) | Multimedia data processing method and device, electronic equipment and storage medium | |
CN113705681A (en) | Lipstick number identification method based on machine learning | |
Powar et al. | Skin detection for forensic investigation |
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 | ||
CP02 | Change in the address of a patent holder |
Address after: 610015 floor 13, building 1, No.1268, middle section of Tianfu Avenue, Chengdu high tech Zone, China (Sichuan) pilot Free Trade Zone, Chengdu Patentee after: Chengdu PinGuo Digital Entertainment Ltd. Address before: 610041 C12-16 building, Tianfu Software Park, hi tech Zone, Sichuan, Chengdu Patentee before: Chengdu PinGuo Digital Entertainment Ltd. |
|
CP02 | Change in the address of a patent holder |