CN102915372B - Image search method, Apparatus and system - Google Patents
Image search method, Apparatus and system Download PDFInfo
- Publication number
- CN102915372B CN102915372B CN201210438698.6A CN201210438698A CN102915372B CN 102915372 B CN102915372 B CN 102915372B CN 201210438698 A CN201210438698 A CN 201210438698A CN 102915372 B CN102915372 B CN 102915372B
- Authority
- CN
- China
- Prior art keywords
- image
- profile
- scene
- gray level
- destination object
- 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
Abstract
The invention discloses a kind of image search method, first greyscale transformation and noise reduction process are carried out to the scene image that photographing module is caught, obtain scene gray level image; Then detected target object profile in described scene gray level image; Finally the feature interpretation of the target gray image in destination object profile or target gray image is uploaded to image retrieval server end and carries out image retrieval, accordingly, the invention also discloses a kind of image retrieving apparatus, mobile terminal and image indexing system, before image retrieval, accurately destination object is extracted by contour detecting, effective minimizing image background, on the impact of image searching result, improves retrieval accuracy.
Description
Technical field
The present invention relates to image processing field, particularly relate to a kind of image search method, Apparatus and system.
Background technology
Image retrieval is divided into based on text and content-based two kinds.At present, the image retrieval that most of search engine provides is all text based retrieval, namely the keyword based on picture appearance information or artificial mark carries out free word retrieval, text based image retrieval needs image to be associated with text, the result for retrieval of image is obtained by text search, its retrieval accuracy affects larger by key word, the visual signature (color or texture etc.) comprised in image often cannot carry out objective description with text, and the difference of subjective understanding will cause the mismatch error in image retrieval.The foundation of database simultaneously needs a large amount of manual operations, and cost is higher.
Along with the appearance of large scale digital image library, text based retrieval cannot adaption demand, CBIR technology (content-basedimageretrieval) is arisen at the historic moment, be different from the way of in original system, image being carried out to artificial mark, content-based retrieval technology extracts the vision content feature of every width image automatically as its index, as color, texture, shape etc.
Along with the development of development of Mobile Internet technology, CBIR is also employed in mobile terminal gradually.Existing mobile image retrieval, all the scene image of mobile terminal photographing module collection is directly uploaded to retrieval server carry out image retrieval, this mode not only data volume is large, higher to network bandwidth requirement, and more seriously photographing module gather destination object image time, often collect the background content near destination object, and image background content can form severe jamming in image retrieval procedure, thus have a strong impact on the accuracy of image retrieval.
Summary of the invention
The object of this invention is to provide a kind of image search method, Apparatus and system, before image retrieval, carry out contour detecting, extract destination object profile, and the target gray image interception in destination object profile is out uploaded to image retrieval server end carries out CBIR, reduce image background to the impact of image searching result, improve retrieval accuracy, the present invention is specially adapted to mobile cloud field of image search.
In order to realize foregoing invention object, the invention provides a kind of image search method, comprising: greyscale transformation and noise reduction process are carried out to the scene image that photographing module is caught, obtains scene gray level image, detected target object profile in described scene gray level image, and the target gray image interception in destination object profile is out uploaded to image retrieval server end carries out image retrieval, or feature detection is carried out to the target gray image in destination object profile obtain feature interpretation, the feature interpretation of the target gray image obtained is uploaded to image retrieval server end and carries out image retrieval, this place feature detection can adopt SIFT, SURF, Ferns, ORB scheduling algorithm, the feature detection such as SIFT can use the CPU (CentralProcessingUnit of mobile terminal, central processing unit) or GPU (GraphicProcessingUnit, graphic process unit) or special asic chip (ASIC, ApplicationSpecificIntegratedCircuit) come.
When the data uploaded to image retrieval server end are target gray image, first image retrieval server end will carry out feature detection to the described target gray image uploaded, obtain the feature interpretation of target gray image, and then the feature interpretation of described target gray image is mated with the feature interpretation of sample image in database, return the resource information corresponding to sample image that the match is successful.And when the data uploaded to image retrieval server end are the feature interpretation of target gray image, image retrieval server end can directly enter images match step.
Preferably, described in described scene gray level image detected target object profile step comprise further: contour detecting is carried out to described scene gray level image, obtains all outline datas in scene gray level image; Closure detection and polygon approach are carried out to described all outline datas, extracts all closed outline consistent with the contour shape of preset template image or extract all closed outlines meeting preset shapes parameter; The largest contours comprising described scene gray level image central point in the described closed outline extracted or minimized profile are defined as destination object profile.
Preferably, described in described scene gray level image detected target object profile step comprise further: contour detecting is carried out to described scene gray level image, obtains all outline datas in scene gray level image; Closure detection and central point detection are carried out to described all outline datas, extracts all closure profiles comprising described scene gray level image central point; Polygon approach is carried out to the described all closure profiles comprising central point extracted, extracts the standard target object outline consistent with the contour shape of preset template image or extract the standard target object outline meeting preset shapes parameter; Largest contours in the described standard target object outline extracted or minimized profile are defined as destination object profile.
Preferably, described one of carrying out described scene gray level image in contour detecting employing Canny operator, Sobel operator, Prewitt operator, Robert operator and Laplacian operator, these algorithms all belong to prior art, are not described in detail at this.
Preferably, after destination object contour detecting out, also comprise: detect and whether have trigger in destination object profile certain limit; When trigger being detected in preset time, then carry out the sectional drawing of target gray image or carry out target image characteristics detecting step, and then initiating image retrieval request to image retrieval server end; And when can't detect trigger in preset time, then terminate this image retrieval flow process.
Preferably, in described scene gray level image before detected target object profile, also comprise: detect in described scene gray level image whether have trigger; When trigger being detected in preset time, then detected target object profile in described scene gray level image, otherwise terminate this image retrieval flow process.
Accordingly, present invention also offers a kind of image retrieving apparatus, comprise greyscale transformation module, contours extract mould and retrieval request module, described image retrieving apparatus is connected with the photographing module of mobile terminal, wherein: described greyscale transformation module, scene image for catching photographing module carries out greyscale transformation and noise reduction process, obtains scene gray level image; Described profile extraction module, for detected target object profile in described scene gray level image; Described retrieval request module, carries out image retrieval for the target gray image interception in destination object profile is out uploaded to image retrieval server end; Or feature detection is carried out to the target gray image in destination object profile, the feature interpretation obtaining target gray image is uploaded to image retrieval server end and carries out image retrieval.
Preferably, described profile extraction module comprises: outline data detecting unit, carries out contour detecting to the scene gray level image that described greyscale transformation module converts obtains, and obtains all outline datas in scene gray level image; Profile closure detects and polygon approach unit, closure detection and polygon approach are carried out to all outline datas that described outline data detecting unit detects, extracts all closed outline consistent with the contour shape of the preset template image that image retrieval server end stores or extract all closed outlines meeting preset shapes parameter; Destination object outline specifying unit, is defined as destination object profile by the largest contours comprising described scene gray level image central point in the described closed outline extracted or minimized profile.
Preferably, described profile extraction module comprises: outline data detecting unit, carries out contour detecting to the scene gray level image that described greyscale transformation module converts obtains, and obtains all outline datas in scene gray level image; Profile closure detects and central point detecting unit, above-mentioned outline data detecting unit detect in all outline datas, extract all closure profiles comprising described scene gray level image central point; Polygon approach unit, polygon approach is carried out to the described all closure profiles comprising central point extracted, extracts the standard target object outline consistent with the contour shape of preset template image or extract the standard target object outline meeting preset shapes parameter; Destination object outline specifying unit, is defined as destination object profile by the largest contours in the described standard target object outline extracted or minimized profile.
Preferably, described image retrieving apparatus also comprises trigger detection module, described trigger detection module is connected between described profile extraction module and described retrieval request module, for after destination object contour detecting out, detect and whether have trigger in destination object profile certain limit; When trigger being detected in preset time, then trigger described retrieval request module work.
Preferably, described image retrieving apparatus also comprises trigger detection module, described trigger detection module is connected between described greyscale transformation module and described profile extraction module, for before described profile extraction module work, detect in scene gray level image whether have trigger; When trigger being detected in preset time, then trigger described profile extraction module work.
Accordingly, present invention also offers a kind of mobile terminal, the image retrieving apparatus comprising photographing module and be connected with photographing module, described image retrieving apparatus is aforesaid image retrieving apparatus.
Accordingly, present invention also offers a kind of image indexing system, comprise mobile terminal and image retrieval server end; Described mobile terminal is above-mentioned mobile terminal, and the scene image that photographing module is caught by described mobile terminal carries out greyscale transformation and noise reduction process, obtains scene gray level image; And destination object profile is detected in described scene gray level image, the target gray image interception in destination object profile is out uploaded to image retrieval server end; After described image retrieval received server-side to described target gray image, feature detection (this place feature detection can adopt SIFT, SURF, Ferns, ORB, BRIEF, BRISK, FREAK scheduling algorithm) is carried out to described target gray image, obtains the feature interpretation of target gray image; And the feature interpretation of described target gray image is mated with the feature interpretation of sample image in database, return the resource information corresponding to the sample image that the match is successful to described mobile terminal.
Accordingly, present invention also offers another kind of image indexing system, comprise mobile terminal and image retrieval server end; Described mobile terminal is above-mentioned mobile terminal, and the scene image that photographing module is caught by described mobile terminal carries out greyscale transformation and noise reduction process, obtains scene gray level image; And destination object profile is detected in described scene gray level image, feature detection is carried out to the target gray image in destination object profile, the feature interpretation obtaining target gray image is uploaded to image retrieval server end, this place feature detection can adopt SIFT, SURF, Ferns, ORB, BRIEF, BRISK, FREAK scheduling algorithm, and the feature detection such as SIFT can have been come with CPU or GPU of mobile terminal or special asic chip; After described image retrieval received server-side to the feature interpretation of described target gray image, the feature interpretation of described target gray image is mated with the feature interpretation of sample image in database, returns the resource information corresponding to the sample image that the match is successful to described mobile terminal.
The present invention has following beneficial effect:
1, the present invention is CBIR, and relative to text based image retrieval mode, its accuracy can not be subject to the restriction of search key;
2, the present invention is by destination object contours extract, removes the background content in photographing module capturing scenes image, effectively can prevent interference, improve image retrieval accuracy;
3, the present invention uploads to the data of image retrieval server end is the subimage of the original image that photographing module is caught or the feature interpretation of subimage, and data volume is relatively little, effectively can shorten uplink time, reduces the network flow consumption of user.
Accompanying drawing explanation
In order to be illustrated more clearly in the embodiment of the present invention or technical scheme of the prior art, be briefly described to the accompanying drawing used required in embodiment or description of the prior art below, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings:
Fig. 1 is embodiment of the present invention image search method process flow diagram;
Fig. 2 is the schematic flow sheet of detected target object profile in scene gray level image in embodiment of the present invention method;
Fig. 3 is embodiment of the present invention image indexing system workflow one;
Fig. 4 is embodiment of the present invention image indexing system workflow two;
Fig. 5 is embodiment of the present invention image retrieving apparatus structural representation one;
Fig. 6 is embodiment of the present invention image retrieving apparatus structural representation two;
Fig. 7 is the profile extraction module structural representation one in the embodiment of the present invention in image retrieving apparatus;
Fig. 8 is the profile extraction module structural representation two in the embodiment of the present invention in image retrieving apparatus.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, be clearly and completely described the technical scheme in the embodiment of the present invention, obviously, described embodiment is only the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not making the every other embodiment obtained under creative work prerequisite, belong to the scope of protection of the invention.
The present invention relates generally to CBIR, image characteristics extraction with mate before, objective contour detection is carried out to input picture, object and background region segmentation is come, then the feature interpretation of the gray level image of destination object or destination object gray level image is uploaded to image retrieval server and carries out image retrieval, reduce image background to the impact of image searching result, improve retrieval accuracy.
In the embodiment of the present invention, CBIR, image characteristics extraction can adopt arbitrary characteristics extraction algorithm to carry out, such as: based on the feature SIFT (Scale-invariantfeaturetransform described, scale invariant feature is changed), SURF (Speeded-UpRobustFeatures, quick robustness feature), ORB (OrientedFASTandRotatedBRIEF), BRIEF (BinaryRobustIndependentElementaryFeatures), BRISK (BinaryRobustInvariantScalableKeypoints), FREAK (FastRetinaKeypoint) etc., for another example: based on the feature Ferns of sorter, wherein Ferns algorithm can with reference to " FastKeypointRecognitionusingRandomFernsM.
m.Calonder, V.Lepetit, P.FuaIEEETransactionsonPatternAnalysisandMachineIntellig ence, Vol.32, Nr.3, pp.448-461, March2010 ".
For convenience of description, in embodiment below, unification selects SIFT algorithm to carry out citing to describe, and this measure can not be considered as the restriction to image characteristics extraction algorithm.
See Fig. 1, be a kind of process flow diagram of embodiment of the present invention image search method, described image search method, comprising:
Step S101: greyscale transformation and noise reduction process are carried out to the scene image that photographing module is caught, obtains scene gray level image;
Step S102: detected target object profile in described scene gray level image;
Step S103: the target gray image interception in destination object profile is out uploaded to image retrieval server end and carries out image retrieval; Or SIFT feature detection is carried out to the target gray image in destination object profile, the description of the SIFT feature of the target gray image obtained is uploaded to image retrieval server end and carries out image retrieval.
In step s 102, the flow process of detected target object profile has two kinds, and two kinds of flow processs include the steps such as contour detecting, the detection of profile closure, polygon approach, central point judgement, and just the sequencing of these steps is different.
First composition graphs 2 is introduced the testing process of the first destination object profile below, mainly comprises following three steps:
A1: contour detecting (rim detection), namely contour detecting is carried out to the scene gray level image obtained in step S101, obtain all outline datas in scene gray level image, in this step, contour detecting can adopt the one in the edge detection operators such as Canny operator, Sobel operator, Prewitt operator, Robert operator and Laplacian.
A2: profile closure detects and polygon approach, namely closure detection and polygon approach are carried out to all outline datas that A1 step detects, extract all closed outlines consistent with the contour shape of preset template image or extract all closed outlines meeting preset shapes parameter.
Wherein, it is check whether profile is closed that closure detects, and gets rid of isolated curved profile to the impact of subsequent detection.And outline polygon matching, there are two kinds of modes, a kind of be in advance with the setting of the form of parameter to extract which kind of or which plant the polygon of shape, when polygon approach, by meeting the closed polygon contours extract of preset shapes parameter request, out (this kind of mode is applicable to application-specific scene, what namely need extraction is regular polygon profile, as the simple polygons such as triangle, quadrilateral or pentagon wherein one or more, install beforehand Rule of judgment, only judges polygon limit number); Another puts into template image in a database (can be one, also can be multiple), then by abstract for the shape of template image be polygon, extract the closed many profile consistent with template image shape, this kind of mode is applicable to arbitrarily shaped image profile, especially the matching of complicated shape image outline can be solved, as facial contour etc.; As in Fig. 2, the template image profile in database is quadrilateral, then, during polygon approach, only extract quadrangular configuration.Certainly in complex situations, two kinds of polygon approach modes can be used in combination.
A3: profile uniqueness judges, the largest contours comprising described scene gray level image central point in the closed outline extracted described in being about to or minimized profile are defined as destination object profile.
Due to profile consistent with template image shape in scene image or to meet the profile of preset shapes parameter not unique, therefore the closed outline extracted through A2 step is not unique most probably yet, in order to obtain destination object profile, need the profile extracted A2 step to carry out a uniqueness to judge, determine destination object profile.Such as, the largest contours that comprises scene gray level image central point can be set or minimized profile is defined as destination object profile, because according to the use habit of general user, for when taking pictures to destination object, destination object can occupy the zone line of screen and relative size is larger, therefore can set the largest contours comprising screen center's point be destination object profile (minimized profile also can, but not too meet general user custom).
In the testing process of the second destination object profile, first carry out central point detection and carry out polygon approach again, detected by central point and can reject most of uncorrelated profile, the testing process (i.e. Fig. 2 flow process) comparing the first destination object profile more easily obtains destination object profile, and concrete steps are as follows:
B1: carry out contour detecting to described scene gray level image, obtains all outline datas in scene gray level image;
B2: closure detection and central point detection are carried out to described all outline datas, extracts all closure profiles comprising described scene gray level image central point;
B3: polygon approach is carried out to the described all closure profiles comprising central point extracted, extracts the standard target object outline consistent with the contour shape of preset template image or extract the standard target object outline meeting preset shapes parameter;
B4: the largest contours in the described standard target object outline extracted or minimized profile are defined as destination object profile.
In step s 103, when the data uploaded to image retrieval server end are target gray image, after image retrieval received server-side image retrieval request, first SIFT feature detection is carried out to the described target gray image uploaded, the SIFT feature obtaining target gray image describes, then the SIFT feature of described target gray image is described and mate with the feature interpretation of sample image in database, return the resource information corresponding to sample image that the match is successful, in this kind of mode, when the testing process of destination object profile adopts Fig. 2 mode, then the whole image retrieval flow process of the embodiment of the present invention can see Fig. 3.
And ought be in step s 103, the data uploaded to image retrieval server end are that the SIFT feature of target gray image is when describing, after image retrieval received server-side image retrieval request, directly carry out images match, in this kind of mode, when the testing process of destination object profile adopts Fig. 2 mode, then the whole image retrieval flow process of the embodiment of the present invention can see Fig. 4.
Preferably, between step S102 and S103, a trigger detecting step can also be increased, namely after destination object contour detecting out, not enter step S103 immediately, but detect whether have trigger in destination object profile certain limit, only in preset time, around destination object profile, detect trigger, just enter S103 step, and then initiate image retrieval request to image retrieval server end; And when can't detect trigger in preset time, then terminate this image retrieval flow process, after this image retrieval flow process terminates, can set and get back to step S101 and repeat searching step, also can be set as terminating whole flow process no longer capturing scenes image.Described destination object profile certain limit is preset as required, such as, can be set to profile to extension 1 to 2 centimetre or inwardly prolong the scope of 1-2 centimetre.
Described trigger can be any preset identifications, a such as LOGO, an or special pattern etc., and this trigger plays the effect of a gauge tap, can realize only uploading the image with trigger to server end, improves retrieval rate.Such as: when the image search method of the embodiment of the present invention is applied to augmented reality, a certain trigger mark can be marked in the picture outline perimeter carrying out augmented reality, the picture that only marked this trigger is just uploaded to image retrieval server retrieves, to realize accurately pushing augmented reality information, avoid user to catch arbitrarily image by terminal to be all sent to server end and to retrieve, the server-side retrieval pressure caused is large, and the possibility of arbitrary image warehouse-in augmented reality is also smaller, duration can be caused to retrieve less than result, reduce Consumer's Experience.
In one embodiment, above-mentioned trigger detecting step can be arranged between step S101 and step S102, trigger whether carry out contour detecting with trigger, realize only carrying out contour detecting and subsequent step to the scene gray level image with trigger, namely whether trigger is had in the scene gray level image that detection S101 step obtains, when trigger being detected in preset time, then start detected target object profile in described scene gray level image, otherwise terminate this image retrieval flow process, greatly can reduce the expense in mobile terminal like this.
See Fig. 5, for the structural representation of embodiment of the present invention image retrieving apparatus, comprise greyscale transformation module 1, contours extract mould 2 and retrieval request module 3, it is (actual when implementing that described image retrieving apparatus is installed in mobile terminal, image retrieving apparatus can the mode of client be installed in mobile terminal), be connected with the photographing module signal of mobile terminal, wherein: described greyscale transformation module 1, scene image for catching photographing module carries out greyscale transformation and noise reduction process, obtains scene gray level image; Described profile extraction module 2, for detected target object profile in described scene gray level image; Described retrieval request module 3, carries out image retrieval for the target gray image interception in destination object profile is out uploaded to image retrieval server end; Or SIFT feature detection is carried out to the target gray image in destination object profile, the SIFT feature description obtaining target gray image is uploaded to image retrieval server end and carries out image retrieval.
Preferably, described image retrieving apparatus also comprises trigger detection module 4, see Fig. 6, described trigger detection module 4 is connected between described profile extraction module 2 and described retrieval request module 3, for after destination object contour detecting out, detect and whether have trigger in destination object profile certain limit; When trigger being detected in preset time, then triggering described retrieval request module work, playing a gauge tap effect.
Preferably, in practice process, trigger detection module 4 can also be connected between greyscale transformation module 1 and profile extraction module 2, for before described profile extraction module work, detect in scene gray level image and whether have trigger, when trigger being detected in preset time, then trigger described profile extraction module 2 and work.
Wherein, profile extraction module 2 has two kinds of structures, respectively can see Fig. 7, Fig. 8.
First composition graphs 7 introduces the first structure of profile extraction module below, and see Fig. 7, described profile extraction module 2 comprises: outline data detecting unit 21, profile closure detect and polygon approach unit 22 and destination object outline specifying unit 23, wherein:
Described outline data detecting unit 21, carries out contour detecting for the scene gray level image be converted to described greyscale transformation module 1, obtains all outline datas in scene gray level image;
Described profile closure detects and polygon approach unit 22, carries out closure detection and polygon approach, extract the polygon meeting matching requirement for all outline datas detected described outline data detecting unit.Wherein, it is check whether profile is closed that closure detects, and gets rid of isolated curved profile to the impact of subsequent detection.And outline polygon matching, there are two kinds of modes, a kind of be in advance with the setting of the form of parameter to extract which kind of or which plant the polygon of shape, when polygon approach, by meeting the closed polygon contours extract of preset shapes parameter request, out (this kind of mode is applicable to application-specific scene, what namely need extraction is regular polygon profile, as triangle, the simple polygon such as quadrilateral or pentagon wherein one or more, can be Rule of judgment as preset shapes preset parameter using polygon limit number, only judge polygon limit number), in this kind of mode, need detect in profile closure and pre-set the polygon parameter needing matching in polygon approach unit 22, another puts into template image in a database (can be one, also can be multiple), then by abstract for the shape of template image be polygon, extract the closed many profile consistent with template image shape, this kind of mode is applicable to arbitrarily shaped image profile, especially the matching of complicated shape image outline can be solved, as facial contour etc., in this kind of mode, need to pre-deposit template image in the database of server end, the detection of profile closure and polygon approach unit 22 are when carrying out polygon approach, need the template image in invoking server client database, carry out outline polygon matching, as in Fig. 2, the template image profile in database is quadrilateral, then, during polygon approach, only extract quadrangular configuration.Certainly in complex situations, two kinds of polygon approach modes can be used in combination.
Described destination object outline specifying unit 23, the largest contours comprising described scene gray level image central point in the described closed outline extracted or minimized profile are defined as destination object profile (during practice, according to user's use habit, generally arranging the largest contours comprising central point is destination object profile).
First composition graphs 8 introduces the second structure of profile extraction module below, see Fig. 8, described profile extraction module 2 comprises: outline data detecting unit 211, profile closure detect and central point detecting unit 222, polygon approach unit 233 and destination object outline specifying unit 234, wherein:
Described outline data detecting unit 211, carries out contour detecting to the scene gray level image that described greyscale transformation module converts obtains, and obtains all outline datas in scene gray level image;
Described profile closure detects and central point detecting unit 222, above-mentioned outline data detecting unit 211 detect in all outline datas, extract all closure profiles comprising described scene gray level image central point;
Described polygon approach unit 233, polygon approach is carried out to the described all closure profiles comprising central point extracted, extracts the standard target object outline consistent with the contour shape of preset template image or extract the standard target object outline meeting preset shapes parameter;
Described destination object outline specifying unit 234, is defined as destination object profile by the largest contours in the described standard target object outline extracted or minimized profile.
The invention also discloses a kind of mobile terminal, the image retrieving apparatus comprising photographing module and be connected with photographing module, described image retrieving apparatus is the image retrieving apparatus disclosed in the embodiment of the present invention.
The invention also discloses a kind of image indexing system, comprise above-mentioned mobile terminal and image retrieval server end that image retrieving apparatus is installed, introduce the present embodiment image indexing system workflow below in conjunction with Fig. 3, mainly comprise the steps:
C1: mobile terminal photographing module catches the scene image comprising destination object, obtains color scene image; Simple understanding, user with mobile terminal camera facing to needing the destination object carrying out image retrieval, to gather the scene image comprising destination object, (destination object can for specific picture, people, the arbitrary objects such as article), the scene image now gathered, except destination object, comprises the background around destination object toward contact.
C2: the scene image of being caught by photographing module carries out greyscale transformation and noise reduction process, obtains scene gray level image;
C3: carry out contour detecting to scene gray level image, obtains outline datas all in image;
C4: carry out closure detection and polygon approach to all outline datas that C3 step obtains, extracts all closed outline consistent with the contour shape of preset template image or extracts all closed outlines of reservation shape;
C5: carry out uniqueness judgement to the profile that C4 screens, to determine destination object profile;
C6: by the target gray image interception in destination object profile out upload images retrieval server end carry out image retrieval;
C7: image retrieval received server-side image retrieval request, carries out SIFT feature detection to the target gray image uploaded, and the SIFT feature obtaining this image describes;
C8: the SIFT feature of described target gray image is described and mates with the feature interpretation of sample image in database, return the resource information corresponding to the sample image that the match is successful to described mobile terminal.
In above-mentioned steps, C1 ~ C6 completes in the terminal, and C7, C8 step is carried out in image retrieval server end, and mobile terminal is end of out being uploaded onto the server by target gray image interception.
Present invention also offers another kind of image indexing system, comprise mobile terminal and image retrieval server end that image retrieving apparatus is installed equally, its workflow is substantially identical with the workflow of image indexing system in embodiment above, unique difference is exactly that mobile terminal is after detecting destination object profile, it not end of directly the target gray image interception in destination object profile out being uploaded onto the server, but in the terminal feature detection is carried out to the target gray image in destination object profile, as SIFT feature detects, the SIFT feature obtaining target gray image is described and is uploaded to image retrieval server end, after received server-side to retrieval request, without the need to carrying out feature point detection, directly to mate, one of its flow process can see Fig. 4, in the present embodiment, what mobile terminal was uploaded to server end is the feature interpretation of destination object gray level image, it is less that data volume uploads gray-scale map relatively.
Image search method of the present invention, Apparatus and system, before image retrieval, accurately destination object is extracted by contour detecting, effectively can remove the background content in photographing module capturing scenes image, effectively can prevent interference, improve image retrieval accuracy, the data uploading to image retrieval server end due to invention are in addition the subimage of the original image that photographing module is caught or the feature interpretation of subimage, data volume is relatively little, effectively can shorten uplink time, reduce the network flow consumption of user.
All features disclosed in this instructions, or the step in disclosed all methods or process, except mutually exclusive feature and/or step, all can combine by any way.
Arbitrary feature disclosed in this instructions (comprising any accessory claim, summary and accompanying drawing), unless specifically stated otherwise, all can be replaced by other equivalences or the alternative features with similar object.That is, unless specifically stated otherwise, each feature is an example in a series of equivalence or similar characteristics.
The present invention is not limited to aforesaid embodiment.The present invention expands to any new feature of disclosing in this manual or any combination newly, and the step of the arbitrary new method disclosed or process or any combination newly.
Claims (9)
1. an image search method, is characterized in that, comprising:
Greyscale transformation and noise reduction process are carried out to the scene image that photographing module is caught, obtains scene gray level image;
Contour detecting is carried out to described scene gray level image, obtains all outline datas in scene gray level image;
Closure detection and central point detection are carried out to described all outline datas, extracts all closure profiles comprising described scene gray level image central point;
Polygon approach is carried out to the described all closure profiles comprising central point extracted, extracts the standard target object outline consistent with the contour shape of preset template image;
Largest contours in the described standard target object outline extracted or minimized profile are defined as destination object profile;
Target gray image interception in destination object profile is out uploaded to image retrieval server end and carries out image retrieval, or
Feature detection is carried out to the target gray image in destination object profile, the feature interpretation of the target gray image obtained is uploaded to image retrieval server end and carries out image retrieval.
2. the method for claim 1, is characterized in that, described one of carrying out described scene gray level image in contour detecting employing Canny operator, Sobel operator, Prewitt operator, Robert operator and Laplacian operator.
3. the method for claim 1, is characterized in that, after destination object contour detecting out, also comprises: detect and whether have trigger in destination object profile certain limit;
When trigger being detected in preset time, then carry out the sectional drawing of target gray image or carry out target image characteristics detecting step, and then initiating image retrieval request to image retrieval server end;
And when can't detect trigger in preset time, then terminate this image retrieval flow process.
4. the method for claim 1, is characterized in that, in described scene gray level image before detected target object profile, also comprises: detect in described scene gray level image whether have trigger;
When trigger being detected in preset time, then detected target object profile in described scene gray level image, otherwise terminate this image retrieval flow process.
5. an image retrieving apparatus, is characterized in that, comprise greyscale transformation module, contours extract mould and retrieval request module, described image retrieving apparatus is connected with the photographing module of mobile terminal, wherein:
Described greyscale transformation module, carries out greyscale transformation and noise reduction process for the scene image of catching photographing module, obtains scene gray level image;
Described profile extraction module, for detected target object profile in described scene gray level image;
Described retrieval request module, carries out image retrieval for the target gray image interception in destination object profile is out uploaded to image retrieval server end; Or feature detection is carried out to the target gray image in destination object profile, the feature interpretation obtaining target gray image is uploaded to image retrieval server end and carries out image retrieval;
Wherein, described profile extraction module comprises:
Outline data detecting unit, carries out contour detecting to the scene gray level image that described greyscale transformation module converts obtains, and obtains all outline datas in scene gray level image;
Profile closure detects and central point detecting unit, above-mentioned outline data detecting unit detect in all outline datas, extract all closure profiles comprising described scene gray level image central point;
Polygon approach unit, carries out polygon approach to the described all closure profiles comprising central point extracted, extracts the standard target object outline consistent with the contour shape of preset template image;
Destination object outline specifying unit, is defined as destination object profile by the largest contours in the described standard target object outline extracted or minimized profile.
6. device as claimed in claim 5, it is characterized in that, described image retrieving apparatus also comprises trigger detection module, described trigger detection module is connected between described profile extraction module and described retrieval request module, for after destination object contour detecting out, detecting and whether have trigger in destination object profile certain limit, when trigger being detected in preset time, then triggering described retrieval request module work; Or
Described trigger detection module is connected between described greyscale transformation module and described profile extraction module, for before described profile extraction module work, detect in scene gray level image and whether have trigger, when trigger being detected in preset time, then trigger described profile extraction module work.
7. a mobile terminal, comprises photographing module, it is characterized in that, described mobile terminal also comprises the image retrieving apparatus be connected with described photographing module, and described image retrieving apparatus is the image retrieving apparatus described in claim 5 or 6.
8. an image indexing system, is characterized in that, comprises mobile terminal and image retrieval server end;
Described mobile terminal is mobile terminal according to claim 7, and the scene image that photographing module is caught by described mobile terminal carries out greyscale transformation and noise reduction process, obtains scene gray level image; And destination object profile is detected in described scene gray level image, the target gray image interception in destination object profile is out uploaded to image retrieval server end;
After described image retrieval received server-side to described target gray image, feature detection is carried out to described target gray image, obtains the feature interpretation of target gray image; And the feature interpretation of described target gray image is mated with the feature interpretation of sample image in database, return the resource information corresponding to the sample image that the match is successful to described mobile terminal.
9. an image indexing system, is characterized in that, comprises mobile terminal and image retrieval server end;
Described mobile terminal is mobile terminal according to claim 7, and the scene image that photographing module is caught by described mobile terminal carries out greyscale transformation and noise reduction process, obtains scene gray level image; And destination object profile is detected in described scene gray level image, feature detection is carried out to the target gray image in destination object profile, the feature interpretation obtaining target gray image is uploaded to image retrieval server end;
After described image retrieval received server-side to the feature interpretation of described target gray image, the feature interpretation of described target gray image is mated with the feature interpretation of sample image in database, returns the resource information corresponding to the sample image that the match is successful to described mobile terminal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210438698.6A CN102915372B (en) | 2012-11-06 | 2012-11-06 | Image search method, Apparatus and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210438698.6A CN102915372B (en) | 2012-11-06 | 2012-11-06 | Image search method, Apparatus and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102915372A CN102915372A (en) | 2013-02-06 |
| CN102915372B true CN102915372B (en) | 2016-02-03 |
Family
ID=47613738
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210438698.6A Active CN102915372B (en) | 2012-11-06 | 2012-11-06 | Image search method, Apparatus and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102915372B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106203361A (en) * | 2016-07-15 | 2016-12-07 | 苏州宾果智能科技有限公司 | A kind of robotic tracking's method and apparatus |
Families Citing this family (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103246742A (en) * | 2013-05-20 | 2013-08-14 | 成都理想境界科技有限公司 | Image retrieval trigger method and augmented reality method |
| KR20150019668A (en) * | 2013-08-14 | 2015-02-25 | 삼성전자주식회사 | Supporting Method For suggesting information associated with search and Electronic Device supporting the same |
| CN103559194B (en) * | 2013-09-12 | 2017-08-08 | 北京智谷睿拓技术服务有限公司 | A kind of searching method, server, system |
| CN103927387B (en) * | 2014-04-30 | 2017-06-16 | 成都理想境界科技有限公司 | Image indexing system and its correlation technique and device |
| CN103984942A (en) * | 2014-05-28 | 2014-08-13 | 深圳市中兴移动通信有限公司 | Object recognition method and mobile terminal |
| CN104077569A (en) * | 2014-06-24 | 2014-10-01 | 纵横壹旅游科技（成都）有限公司 | Image recognizing method and system |
| CN105426899B (en) * | 2014-09-19 | 2019-11-08 | 腾讯科技（北京）有限公司 | Vehicle identification method, device and client |
| CN104850851A (en) * | 2015-04-22 | 2015-08-19 | 福州大学 | ORB feature point matching method with scale invariance |
| CN107153666B (en) * | 2016-03-02 | 2022-03-04 | 魏立江 | Picture searching method |
| CN107767162A (en) * | 2016-08-18 | 2018-03-06 | 深圳市劲嘉数媒科技有限公司 | Advertisement acquisition method and device based on packing box |
| CN106649741A (en) * | 2016-12-26 | 2017-05-10 | 努比亚技术有限公司 | Picture searching method, device and system |
| CN110866137A (en) * | 2018-08-09 | 2020-03-06 | 中兴通讯股份有限公司 | Image processing method, device and storage medium |
| CN109582812A (en) * | 2018-11-21 | 2019-04-05 | 安徽云融信息技术有限公司 | A kind of search method for computer picture |
| CN110099301A (en) * | 2019-04-28 | 2019-08-06 | 深圳回收宝科技有限公司 | The method, apparatus and portable terminal of terminal detection are carried out based on interception image |
| CN110135483A (en) * | 2019-04-30 | 2019-08-16 | 北京百度网讯科技有限公司 | The method, apparatus and relevant device of training image identification model |
| CN110097673A (en) * | 2019-05-17 | 2019-08-06 | 北京深醒科技有限公司 | A kind of gate inhibition's recognition methods based under infrared camera |
| CN112416960A (en) * | 2020-11-18 | 2021-02-26 | 腾讯科技（深圳）有限公司 | Data processing method, device and equipment under multiple scenes and storage medium |
| CN115309988B (en) * | 2022-08-10 | 2023-07-07 | 上海迪塔班克数据科技有限公司 | Webpage search content matching method, system and storage medium |
| CN116012924B (en) * | 2023-01-30 | 2023-06-27 | 人民网股份有限公司 | Face gallery construction method and device and computing equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101789005A (en) * | 2010-01-22 | 2010-07-28 | 深圳创维数字技术股份有限公司 | Image searching method based on region of interest (ROI) |
| CN102622420A (en) * | 2012-02-22 | 2012-08-01 | 哈尔滨工程大学 | Trademark image retrieval method based on color features and shape contexts |
| CN102750535A (en) * | 2012-04-01 | 2012-10-24 | 北京京东世纪贸易有限公司 | Method and system for automatically extracting image foreground |
-
2012
- 2012-11-06 CN CN201210438698.6A patent/CN102915372B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101789005A (en) * | 2010-01-22 | 2010-07-28 | 深圳创维数字技术股份有限公司 | Image searching method based on region of interest (ROI) |
| CN102622420A (en) * | 2012-02-22 | 2012-08-01 | 哈尔滨工程大学 | Trademark image retrieval method based on color features and shape contexts |
| CN102750535A (en) * | 2012-04-01 | 2012-10-24 | 北京京东世纪贸易有限公司 | Method and system for automatically extracting image foreground |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106203361A (en) * | 2016-07-15 | 2016-12-07 | 苏州宾果智能科技有限公司 | A kind of robotic tracking's method and apparatus |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102915372A (en) | 2013-02-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102915372B (en) | Image search method, Apparatus and system | |
| JP6336117B2 (en) | Building height calculation method, apparatus and storage medium | |
| CN109284729B (en) | Method, device and medium for acquiring face recognition model training data based on video | |
| JP4553650B2 (en) | Image group representation method, descriptor derived by representation method, search method, apparatus, computer program, and storage medium | |
| CN107833213B (en) | Weak supervision object detection method based on false-true value self-adaptive method | |
| CN103824053B (en) | The sex mask method and face gender detection method of a kind of facial image | |
| CN105488468B (en) | A kind of localization method and device of target area | |
| KR101322168B1 (en) | Apparatus for real-time face recognition | |
| CN105303150B (en) | Realize the method and system of image procossing | |
| EP3324368A1 (en) | Scenario reconstruction method and apparatus, terminal device, and storage medium | |
| CN106991370B (en) | Pedestrian retrieval method based on color and depth | |
| CN103853724B (en) | multimedia data classification method and device | |
| CN104346630B (en) | A kind of cloud flowers recognition methods of heterogeneous characteristic fusion | |
| JP2004062605A (en) | Scene identification method and device, and program | |
| CN106991448A (en) | A kind of portrait compares processing method | |
| CN110148147A (en) | Image detecting method, device, storage medium and electronic device | |
| CN104951440B (en) | Image processing method and electronic equipment | |
| CN103049911B (en) | Contour detecting stability judging method and image search method | |
| CN110222627A (en) | A kind of face amended record method | |
| CN106096520A (en) | A kind of information-pushing method and device | |
| JP6463095B2 (en) | Image search apparatus and method | |
| CN112784854A (en) | Method, device and equipment for segmenting and extracting clothing color based on mathematical statistics | |
| CN117078624A (en) | Light-weight concrete crack detection method and device based on multidimensional attention module | |
| JP2001167273A5 (en) | ||
| CN105956550B (en) | A kind of method and apparatus that video identifies |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |