CN105683996A - Method for determining local differentiating color for image feature detectors - Google Patents
Method for determining local differentiating color for image feature detectors Download PDFInfo
- Publication number
- CN105683996A CN105683996A CN201380080597.9A CN201380080597A CN105683996A CN 105683996 A CN105683996 A CN 105683996A CN 201380080597 A CN201380080597 A CN 201380080597A CN 105683996 A CN105683996 A CN 105683996A
- Authority
- CN
- China
- Prior art keywords
- image
- response
- vector
- computing equipment
- local difference
- 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
- 238000000034 method Methods 0.000 title claims description 33
- 230000004044 response Effects 0.000 claims abstract description 193
- 239000011159 matrix material Substances 0.000 claims description 50
- 230000006978 adaptation Effects 0.000 claims description 14
- 238000001914 filtration Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 5
- 238000001514 detection method Methods 0.000 abstract description 13
- 238000005516 engineering process Methods 0.000 abstract description 11
- 238000004891 communication Methods 0.000 description 13
- 230000006870 function Effects 0.000 description 6
- 238000010191 image analysis Methods 0.000 description 5
- 238000003860 storage Methods 0.000 description 5
- 238000013500 data storage Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000001737 promoting effect Effects 0.000 description 2
- 230000007480 spreading Effects 0.000 description 2
- 238000003892 spreading Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- YIWGJFPJRAEKMK-UHFFFAOYSA-N 1-(2H-benzotriazol-5-yl)-3-methyl-8-[2-[[3-(trifluoromethoxy)phenyl]methylamino]pyrimidine-5-carbonyl]-1,3,8-triazaspiro[4.5]decane-2,4-dione Chemical compound CN1C(=O)N(c2ccc3n[nH]nc3c2)C2(CCN(CC2)C(=O)c2cnc(NCc3cccc(OC(F)(F)F)c3)nc2)C1=O YIWGJFPJRAEKMK-UHFFFAOYSA-N 0.000 description 1
- 102000002274 Matrix Metalloproteinases Human genes 0.000 description 1
- 108010000684 Matrix Metalloproteinases Proteins 0.000 description 1
- MKYBYDHXWVHEJW-UHFFFAOYSA-N N-[1-oxo-1-(2,4,6,7-tetrahydrotriazolo[4,5-c]pyridin-5-yl)propan-2-yl]-2-[[3-(trifluoromethoxy)phenyl]methylamino]pyrimidine-5-carboxamide Chemical compound O=C(C(C)NC(=O)C=1C=NC(=NC=1)NCC1=CC(=CC=C1)OC(F)(F)F)N1CC2=C(CC1)NN=N2 MKYBYDHXWVHEJW-UHFFFAOYSA-N 0.000 description 1
- NIPNSKYNPDTRPC-UHFFFAOYSA-N N-[2-oxo-2-(2,4,6,7-tetrahydrotriazolo[4,5-c]pyridin-5-yl)ethyl]-2-[[3-(trifluoromethoxy)phenyl]methylamino]pyrimidine-5-carboxamide Chemical compound O=C(CNC(=O)C=1C=NC(=NC=1)NCC1=CC(=CC=C1)OC(F)(F)F)N1CC2=C(CC1)NN=N2 NIPNSKYNPDTRPC-UHFFFAOYSA-N 0.000 description 1
- VCUFZILGIRCDQQ-KRWDZBQOSA-N N-[[(5S)-2-oxo-3-(2-oxo-3H-1,3-benzoxazol-6-yl)-1,3-oxazolidin-5-yl]methyl]-2-[[3-(trifluoromethoxy)phenyl]methylamino]pyrimidine-5-carboxamide Chemical compound O=C1O[C@H](CN1C1=CC2=C(NC(O2)=O)C=C1)CNC(=O)C=1C=NC(=NC=1)NCC1=CC(=CC=C1)OC(F)(F)F VCUFZILGIRCDQQ-KRWDZBQOSA-N 0.000 description 1
- JAWMENYCRQKKJY-UHFFFAOYSA-N [3-(2,4,6,7-tetrahydrotriazolo[4,5-c]pyridin-5-ylmethyl)-1-oxa-2,8-diazaspiro[4.5]dec-2-en-8-yl]-[2-[[3-(trifluoromethoxy)phenyl]methylamino]pyrimidin-5-yl]methanone Chemical compound N1N=NC=2CN(CCC=21)CC1=NOC2(C1)CCN(CC2)C(=O)C=1C=NC(=NC=1)NCC1=CC(=CC=C1)OC(F)(F)F JAWMENYCRQKKJY-UHFFFAOYSA-N 0.000 description 1
- 230000008649 adaptation response Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000000151 deposition Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- SFMJNHNUOVADRW-UHFFFAOYSA-N n-[5-[9-[4-(methanesulfonamido)phenyl]-2-oxobenzo[h][1,6]naphthyridin-1-yl]-2-methylphenyl]prop-2-enamide Chemical compound C1=C(NC(=O)C=C)C(C)=CC=C1N1C(=O)C=CC2=C1C1=CC(C=3C=CC(NS(C)(=O)=O)=CC=3)=CC=C1N=C2 SFMJNHNUOVADRW-UHFFFAOYSA-N 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000001615 p wave Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012887 quadratic function Methods 0.000 description 1
- 238000000985 reflectance spectrum Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/469—Contour-based spatial representations, e.g. vector-coding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
Abstract
Technologies for multi-channel feature detection include a computing device to determine a filter response of each image channel of a multi-channel image for one or more image filters. The computing device determines a local differentiating color vector based on the filter responses, applies the filter responses to the local differentiating color vector to generate an adapted response, and determines a total response of the multi-channel image based on the adapted response.
Description
Background technology
Computer vision utilizes multiple image feature detector to carry out " point-of-interest " in feature or the image of recognition image.Image feature detector can be identified according to special algorithm/detector edge, angle point, spot (, the point of interest of analyzed imageRegion) and/or crestal line. For example, Canny algorithm and Sobel wave filter are carried out rim detection; Harris detector is carried out angle point inspectionSurvey; And the angle in Laplce-Gauss (LoG), Hai Sen-Gauss determinant and difference of Gaussian (DoG) detector recognition imagePoint and spot. Feature detection system usually utilizes the combination of algorithm and detector more accurately to identify the feature of analyzed image.
For example accelerate robust features (SURF), the conversion of yardstick invariant features (SIFT), Canny, Harris and Sobel etc. altogetherDetect and describe the feature of single channel image (, gray level image) with property detector. Therefore, divide as the preliminary of feature detectionAnalyse step, multichannel image (, coloured image) must be transformed into single channel image, and it can cause obvious image information to be lostLose. For example, the image pixel value of single channel gray level image can be used as each respective pixel value in the road of multichannel imageLinear combination and generating. So, there is completely different color but there is the multichannel image picture that identical single channel gray scale representsContrast between element is because greyscale transformation loses. For example, although some algorithms utilize color model based on perception (,CSIFT uses Kubelka-Munk theory, and it makes the reflectance spectrum modelling of color body), they use global color to ashDegree mapping, it causes information dropout.
Brief description of the drawings
By example, unrestriced mode illustrates concept described herein in the accompanying drawings. For illustrated simple and clearFor the purpose of, illustrated in the drawings element is not necessarily drawn in proportion. In the place of thinking fit, label repeats to indicate among figureCorrespondence or like.
Fig. 1 is the simplified block diagram of at least one embodiment of the computing equipment for carrying out multi-channel feature detection;
Fig. 2 is the simplified block diagram of at least one embodiment of the environment of the computing equipment of Fig. 1;
Fig. 3 is the simplification stream of carrying out at least one embodiment of the method for multi-channel feature detection for the computing equipment to Fig. 1Cheng Tu;
Fig. 4 is the letter of at least one embodiment of the method for the local difference color vector on the computing equipment for determining Fig. 1Change flow chart;
Fig. 5 and 6 is respectively the method that detects based on the multi-channel feature of Fig. 3 and catching as the SURF property detector of kernelThe figure of the point of interest of image and its identification.
Detailed description of the invention
Although concept of the present disclosure is subject to various amendments and alternative impact, its specific embodiment has passed through in figureExample illustrates and will describe in detail herein. But, should be appreciated that concept of the present disclosure is not limited to disclosed specificThe intention of form, and contrary, is intended that and will contains all modifications consistent with the claim of enclosing with the disclosure, be equal toThing and alternate item.
The reality that instruction is described of quoting to " embodiment ", " embodiment ", " illustrative embodiment " etc. in descriptionExecute example and can comprise special characteristic, structure or characteristic, but each embodiment can comprise or can not necessarily comprise this special characteristic, structureOr characteristic. In addition such phrase identical embodiment of definiteness that differs. In addition, when special characteristic, structure or characteristic are together with enforcementWhen example is described, think that the embodiment no matter whether clearly describing together with other realizes such feature, structure or characteristic, this is at thisIn field in technical staff's knowledge. In addition, should recognize that employing " at least one A, B and C " form is included in the item in listCan mean: (A); (B); (C); (A and B); (B and C); Or (A, B and C). Similarly, adopt " at least one in A, B or C "The item that form is listed can mean: (A); (B); (C); (A and B); (B and C); Or (A, B and C).
Disclosed embodiment can adopt hardware, firmware, software or its any combination to realize in some cases. DisclosedEmbodiment for example also can be embodied as, by one or more temporary or nonvolatile machine readable (, computer-readable) storage mediumsCarry or the instruction of storage thereon, these instructions can be read and be carried out by one or more processors. Machine readable storageMedium can be presented as any memory device, mechanism or other physics for adopting machine-readable form storage or the information of transmissionStructure (for example, volatibility or nonvolatile memory, media disc or other media devices).
In the drawings, some structures or method characteristic can adopt specific setting and/or sequence to illustrate. But should recognize canDo not need such specific setting and/or sequence. On the contrary, in certain embodiments, such feature can adopt with illustrativeDifferent mode shown in figure and/or order arrange. In addition, comprise that in specific pattern structure or method characteristic are not intended to secretlyReferring to needs such feature in all embodiments, and in certain embodiments, can not comprise that such feature or its can be withOther Feature Combinations.
With reference now to Fig. 1,, the computing equipment 100 detecting for multi-channel feature is configured to detect the feature of multichannel image(for example, the point of interest such as such as angle point, edge, spot). In order to do like this, computing equipment 100 is in recognition image characteristic aspect profitUse information instead of single channel or gray level image (for example, rear changing image) from multiple image channels. In illustrative enforcementIn example, computing equipment 100 is configured to realize low complex degree noniterative algorithm for calculating local difference color (LDC) vector,Wherein the receptance function of kernel can be expressed as linearity or Quadratic Function Optimization. Should recognize that such situation contains single-pass widelyRoad property detector, it has can be suitable for the kernel that uses together with LDC vector. For example, can calculate the filter of second order space derivationRipple device response Dxx、DxyAnd Dyy, wherein x and y are the space coordinates of image. So, the response of LoG kernel can adopt linear formsExpress. The response of SURF kernel can be expressed as quadratic form,
The response of original Harris kernel can be expressed as quadratic form,
Wherein k is algorithm parameter. In addition, the response squared of Canny kernel (squareresponse) can be expressed as quadratic form, wherein DxAnd DyBe first order spatial derivative filter response, wherein x and y are stillThe space coordinates of image.
Computing equipment 100 can be presented as any type that can multi-channel feature detects and carry out function described hereinComputing equipment. For example, computing equipment 100 can be presented as cell phone, smart phone, tablet PC, net book, notebook,Ultrabook, laptop computer, personal digital assistant, mobile internet device, desktop computer, mixing apparatus and/orAny other calculating/communication equipment. As illustrated in fig. 1, illustrative computing equipment 100 comprises processor 110, input/defeatedGo out (" I/O ") subsystem 112, memory 114, data storage 116, telecommunication circuit 118 and one or more ancillary equipment 120.In addition, ancillary equipment 120 comprises filming apparatus 122 and display 124. Certainly, in other embodiments, computing equipment 100 canComprise other or additional components, that for example conventionally for example, find in typical calculation equipment (, various input-output apparatus)A bit. In addition, in certain embodiments, one or more in illustrative components be included in another parts or in addition fromA part for another parts. For example, in certain embodiments, memory 114 or its part can be included in processor 110.
Processor 110 can be presented as the processor of any type that can carry out function described herein. For example, processorCan be presented as monokaryon or polycaryon processor, digital signal processor, microcontroller or other processors or processing/control circuit.Similarly, memory 114 can be presented as volatibility or non-volatile the depositing of any type that can carry out function described hereinReservoir or data storage. In operation, memory 114 can store computing equipment 100 operating period use various data andSoftware, for example operating system, application, program, storehouse and driver. Memory 114 is coupled in communicatedly via I/O subsystem 112Processor 110, this I/O subsystem 110 can be presented as for promoting its of processor 110, memory 114 and computing equipment 100The circuit of the input/output operations of his parts and/or parts. For example, I/O subsystem 112 can be presented as or comprise in addition storageDevice controller maincenter, I/O control axis, firmware device, communication link (, point-to-point link, bus links, wire,Cable, photoconduction, printed circuit board trace, etc.) and/or for promoting miscellaneous part and the subsystem of input/output operations. OneIn a little embodiment, I/O subsystem 112 can form a part for SOC(system on a chip) (SoC) and together with processor 110, memory 114Be included on single integrated circuit chip with together with the miscellaneous part of computing equipment 100.
Data storage device 116 can be presented as for the short-term of data or longer-term storage and the equipment of any type configuringOr multiple equipment, for example, memory devices and circuit, storage card, hard disk drive, solid-state drive or other data storages are establishedStandby. Telecommunication circuit 118 can be presented as and can between computing equipment 100 and other remote equipments, realize by network (not shown)Telecommunication circuit, equipment or its set of any type of communication. In order to do like this, according to for example network type, (it can be presented asCan promote the communication network of any type of communicating by letter between computing equipment 100 and remote equipment), telecommunication circuit 118 can makeFor example, for example, with any applicable communication technology (, wired or wireless communication) and associated agreement (, Ethernet, Bluetooth?、Wi-Fi?, WiMAX, etc.) implement such communication.
The ancillary equipment 120 of computing equipment 100 can comprise extra periphery or the interface equipment of any amount. Ancillary equipment120 particular devices that comprise can be depending on type and/or the desired use of for example mobile computing device 100. As discussed above, ancillary equipment 120 comprises filming apparatus 122 and display 124. Filming apparatus 122 can be presented as for catching appointing of imageWhat periphery or integrated equipment, for example still life filming apparatus, video capture device, IP Camera or can capturing video and/orOther equipment of image. Filming apparatus 122 can be for example for catching the wherein detected multichannel image of feature. Computing equipment100 display 124 can be presented as any or multiple display screen, and information can show to the spectators of computing equipment 100 thereon.Display 124 can be presented as or use in addition any applicable display technology, and it comprises for example liquid crystal display (LCD), sends outOptical diode (LED) display, cathode ray tube (CRT) display, plasma scope and/or other display technologies.Display 124 can be for example for showing the image of the overall response of indicating analyzed image. To calculate to establish although be shown in Figure 1 forStandby 100 integrated part, should recognize in other embodiments, filming apparatus 122 and/or display 124 can be away from calculatingEquipment 100 but communicative couplings are in computing equipment 100.
With reference now to Fig. 2,, in use, computing equipment 100 detects and sets up environment 200 multi-channel feature. Following articles and opinionsState, the wave filter of the indivedual image channels of computing equipment 100 based on local difference color (LDC) vector sum multichannel image ringsShould carry out to determine total image response of analyzed multichannel image. The Illustrative environment 200 of computing equipment 100 comprises image captureModule 202, image analysis module 204, display apparatus module 206 and communication module 208. In addition, image analysis module 204 comprises figurePicture filtration module 210, local difference color module 212 and response determination module 214. Image capture module 202, graphical analysis mouldPiece 204, display apparatus module 206, communication module 208, image filtering module 210, local difference color module 212 and response are determinedEach in module 214 is presented as hardware, software, firmware or its combination. In addition, in certain embodiments, illustrative modulesIn the part that can form another module.
Image capture module 202 is controlled filming apparatus 122 and is caught image (for example, the use in the visual field of filming apparatus 122Detect in multi-channel feature). According to specific embodiment, image can be used as streamcast video or is hunted down as indivedual image/frames. ?In other embodiment, image capture module 202 can be retrieved multichannel image in addition for analyzing and feature detection. For example, manyChannel image can utilize communication module 208 for example, to receive from remote computing device (, in cloud computing environment). Should recognize and catchImage can be presented as any applicable multichannel image. For example, image can be triple channel image, for example RGB(is red-green-Blue), HSL(hue-saturation-brightness) or HSV(tone-saturation degree-numerical value) image. Should further recognize described hereinMultichannel image feature detection can be applicable to the image channel of any type, and it for the passage in achromaticity space (for example, comprisesThe RGB-D(degree of depth), infrared, hygrogram, microwave figure or other image channels).
Image analysis module 204 is retrieved the image of catching with filming apparatus 122 from image capture module 202. IllustrativeIn embodiment, image spreading spatial point is set up to coordinate to image analysis module 204 and parameter (for example, represents for scale-spaceAnd/or use together with scale-space detector). In addition,, as below discussed more in detail, image analysis module 204 is to quiltAnalysis image is applied various wave filters, each picture point (or its subset) to image and is determined LDC vector and definite imageThe overall response (or its subset) of each picture point.
Image filtering module 210 is determined the filter of each image channel of multichannel image to one or more image filtersRipple device response (, the result to image applications image filter). For example, in certain embodiments, image filter can be appliedIn each pixel of image. In doing so, should recognize that image filter for example can use " windowing " method and apply, whereinImage filter is applied to neighborhood of pixels (for example, having the size of image filter kernel). Although image filter generally shouldFor indivedual pixels of image channel, for describe simple and clear for the purpose of, image filter can be described as application hereinIn the value of whole image channel or other structures instead of indivedual pixels. Multichannel image comprises the enforcement of three passages thereinIn example, image filtering module 210 generates to the each image filter of each application in three passages and based on this wave filterCorresponding filter response. It will be appreciated that for the filter response of the specific image passage of multichannel image can be expressed as toAmount, it comprises the correspondence response of image channel to one or more image filters. In addition, vector can be described as correspondence image like this" response vector " of passage or " vector response ". In addition, in certain embodiments, the specific image wave filter of employing must be lineProperty or quadratic form image filter. In other embodiments, LDC vector can be applicable to the pixel of original image passage and does not appointFiltering before what or only there is common/identical wave filter.
Local difference color module 212 is based on determining partial error by the definite filter response of image filtering module 210Divide color vector. As below discussed in detail, the wave filter to image channel is calculated or be defined as to local difference color vectorThe linear combination of response limits weight and generation extreme value (, minimum or maximum, it depends on specific embodiment) overall responseVector. For linear forms, local difference color module 212 determines that local difference color vector is with true to each image channelThe vector of the collinear vector of fixed overall response. For quadratic form, local difference color vector is defined as characteristic vector (or specificationChange characteristic vector), and its extremal eigenvalue corresponding to the symmetrical matrix of specific generation (that is, and maximum or minimal eigenvalue, it depends onIn specific embodiment). So, in illustrative embodiment, local difference color vector can adopt closed form to express, instead of calculatesFor example, for the result of optimized algorithm (, cost function minimized or maximize).
Response determination module 214 is to the local difference face of the image filter response application being generated by image filtering module 210Look vector generates the response of adaptation and the response based on this adaptation determines the overall response of multichannel image. In illustrative realityExecute in example, response determination module 214 is by independently calculating each image channel of local difference color vector and multichannel imageThe dot product of response vector come the local difference color vector of image filter response application. In addition, as below discussed more in detailState, response determination module 214 is by specific filter and/or the parameter of feature detection algorithm and the sound of adaptation based on adoptingShould generate scalar value and determine the overall response of multichannel image.
In illustrative embodiment, response determination module 214 also suppresses the non-extreme value in space of the overall response of multichannel imageResponse. That is, in certain embodiments, response determination module 214 is removed non-point of interest from overall response, and it can be based on predefined threshold valuePredefine. That is to say, point of interest can be identified as have more than threshold value or below the office of (it depends on specific embodiment)The picture point of portion's extreme value response. So, only point of interest is retained in overall response.
Display apparatus module 206 is configured to, on display 124, the user of computing equipment 100 is presented to image for watching.For example, display apparatus module 206 can show the image (referring to Fig. 5) of one or more catching/receive and/or total sound of indicating imageThe image (referring to Fig. 6) of answering. In addition, should recognize that display apparatus module 206 can present in another stage of feature detection processThe visual depiction of image. For example, display apparatus module 206 can suppress to present indivedual filter responses, office before non-extreme value responsePortion's difference color vector, the response of adaptation and/or the figure of overall response and/or text are described.
Communication module 208 is by communicating by letter between network processes computing equipment 100 and remote equipment. As discussed above,Communication module 208 can for example receive multichannel image, for analyzing (, in cloud computing environment or for unloading from remote computing deviceCarry and carry out). So, in certain embodiments, the result (for example, overall response) that communication module 208 also can be analyzed feature detectionBe sent to remote computing device.
With reference now to Fig. 3,, in use, computing equipment 100 executing methods 300 are for carrying out multi-channel feature inspectionSurvey. Illustrative method 300 starts with the frame 302 of Fig. 3, and wherein computing equipment 100 determines whether to carry out multi-channel feature detection. AsFruit computing equipment 100 determines that carrying out multi-channel feature detects, and computing equipment 100 is set up image spreading spatial point in frame 304Coordinate system. That is to say, computing equipment 100 is set up for example cartesian coordinate system (for example, conventional x-and y-axle) and additional parameter(for example, yardstick) to use together with scale-space image feature detector.
In frame 306, computing equipment 100 based on one or more image filters (for example, sea plug determinant, Canny,Sobel wave filter etc.) determine the filter response of each image channel. In doing so, in frame 308, computing equipment 100 basesIn filter response as discussed above, each image channel is generated to response vector (, by indivedual image channels are appliedImage filter). For example, suppose that analyzed multichannel image is triple channel RGB(R-G-B) image and employing GaussThe second order local derviation (, the component of Hesse matrices) of wave filter is as image filter. Therefore, image filter comprises gxx、gyyWithgxy, it is the second order local derviation about correspondence image dimension. In such embodiments, each (, the g in image filterxx、gyyAnd gxyIn each) be applied to red channel come to red image passage generate response vector. As discussed above, imageWave filter can be applicable to each pixel of image. Therefore, can generate response vector to each pixel of image channel. Similarly,Each blue image passage and green image passage of being applied in image filter, makes the each generation response in passageVector. Each response vector can be reduced to scalar value. For example, sea plug determinant can be by quadratic formDetermine, wherein B is predefined matrix and gxx、gyyAnd gxyTo closeThe second order local derviation of the Gaussian filter of taking in corresponding space coordinates x and/or y. Certainly, other embodiment can utilize different numbersThe image filter of amount and/or analysis have the image of varying number passage. So, in the ordinary course of things, suppose to exist n to lead toA road and p wave filter. Under these circumstances, computing equipment 100 generate have size/length p(or, more specifically, big or small p× 1) a n response vector (, one, each passage), wherein element is the filter for the image channel of correspondence image wave filterThe response of ripple device.
In frame 310, the filter response of computing equipment 100 based on each image channel determine local difference color toAmount (for example, normalization LDC vector). That is to say, computing equipment 100 utilizes for the response vector of image channel and generates officePortion's difference color vector. In order to do like this, computing equipment 100 executing methods 400 for determine local difference color vector,As shown in Figure 4. Illustrative method 400 starts with frame 402, and wherein computing equipment 100 determines whether to generate local differenceColor vector. If so, computing equipment 100 in frame 404 to image generate or in addition determine symmetric form matrixA. Should recognize that computing equipment 102 can use any applicable technology, algorithm and/or mechanism to generate the symmetric figure of quadratic form matrixFormula. For example, in one embodiment, in frame 406, computing equipment 100 can calculate the each of A matrix, wherein for image channel i and j,. In such embodiments, qijRepresentative is positioned atThe A entry of a matrix element that i is capable and j is listed as, f is the response vector for the image channel corresponding with the index of f, T is transpositionOperator (that is, fT is the transposition of f), and B is the predefined matrix based on one or more image filters. For example,, aboveIn the embodiment about Hesse matrices describing, B matrix may be defined as:
And be that priori is known. In another embodiment, B matrix can calculate based on image and/or filter parameter. SayingIn bright property embodiment, computing equipment 100 will for example, be calculated as for the matrix A of triple channel image (, RGB image) in frame 408:
Certainly, in other embodiments, analyzed image can comprise still less or the passage of larger quantity, and in such enforcementIn example, matrix A has corresponding size, and (for example, analyzed image comprises that in the embodiment of four passages etc. be 4 × 4 squares thereinBattle array). That is to say, matrix A is presented as n × n matrix, and wherein n is the quantity of image channel.
In frame 410, computing equipment 100 is determined the characteristic value of A matrix. Should recognize that computing equipment 100 can utilize anyApplicable technology, algorithm or mechanism are done like this. For example, A can be determined and utilize to computing equipment 100 in its characteristic value of identificationThe characteristic equation of matrix. In frame 412, the computing equipment 100 identification characteristic vector corresponding with the extremal eigenvalue of A matrix(that is, and maximum or minimal eigenvalue, it depends on specific embodiment) and in frame 414, the spy of computing equipment 100 selective recognitionsLevy vector as local difference color vector. In doing so, in certain embodiments, computing equipment 100 can be given birth in frame 416Be paired in the unit vector of recognition feature vector. , computing equipment 100 can make characteristic vector standardize to generate and characteristic vectorCorresponding unit vector, it is chosen as local difference color vector.
With reference to figure 3, in frame 312, computing equipment 100 is to the local difference color vector application image filter that generates/determineRipple device responds the response of the adaptation that becomes corresponding next life. In doing so, in frame 314, computing equipment 100 computed image wave filtersThe dot product (for example, generating single vector) of response and local difference color vector. For example, the sea plug example of describing all the timeIn, to the local difference color vector of all path computations of image with comprise the vectorial dot product of second order local derviation, it is with vectorTransposition is multiplied by local difference color vector equivalence. Particularly, for all passages of image, willTake advantage ofWith local difference color vector. In frame 316, the response of computing equipment 100 based on adapting to generates overall response. In illustrative realityExecute in example, computing equipment 100 based on adapt to response and the special characteristic detection algorithm/wave filter of use generate scalar value.For example, use therein in the embodiment of Hesse matrices, computing equipment 100 can utilize the parameter of Hesse matrices and/or featureGenerate overall response (for example, using sea plug determinant). Should recognize that computing equipment 100 can utilize any applicable technology, algorithmAnd/or mechanism is done like this.
In frame 318, computing equipment 100 suppresses the non-extreme value response in space of overall response in extending space. , at frameIn 320, computing equipment 100 can be removed non-point of interest from overall response. As discussed above, point of interest and non-point of interest can be based in advanceDefinition threshold value is distinguished. For example, in one embodiment, have and exceed the local extremum response of predefined threshold value or intensity levelThe space diagram picture point of overall response is regarded as " point-of-interest " or " point of interest ", and has the local extremum that does not exceed predefined threshold valueThe space diagram picture point of the overall response of response or intensity level is non-point of interest. Should recognize that computing equipment 100 can utilize and has secondaryAny applicable feature detection algorithm of type receptance function (for example, SURF) and can adopt any applicable mode identify " sensePoint of interest ". As instruction above, according to special algorithm, point-of-interest can comprise angle point, edge, spot and/or other images spyProperty. In addition, in certain embodiments, in frame 316, the generation of overall response comprises that suppressing the non-extreme value in space responds.
As discussed above, computing equipment 100 can generate and show the overall response of the analyzed many channel image of instructionImage watch for user. For example, the analyzed image 500 of simplification is shown in Figure 5, and the example output image of simplifyingIts multi-channel feature based on image 500 of 600(detects (utilizing SURF kernel) and generates illustratively) shown in Figure 6. ?In the output image 600 of simplifying, the point of interest/feature of identification is depicted as and different means to have corresponding difference with shade circleThe circle of color. Certainly, should recognize that image 600 is by the letter of the real world output image that uses technology disclosed herein to generateChange version, and such real world output image can use and has wider different colours and size (it depends onFor example original analysis image) the circle of greater or lesser quantity identify point-of-interest/feature. In addition, should recognize and single-passRoad gray feature detects different, by computing equipment 100, analyzed image is carried out to generate output image as described hereinFeature detection do not stand information loss intrinsic in greyscale transformation.
Example
In the illustrated examples that technology disclosed herein is below provided. The embodiment of technology can comprise in example described belowAny or multiple and any combination.
Example 1 comprises the computing equipment detecting for multi-channel feature, and this computing equipment comprises: image filtering module, useIn the filter response of one or more image filters being determined to each image channel of multichannel image; Local difference colorModule, for determining local difference color vector based on filter response; With response determination module, for (i) to local differenceColor vector filter application respond to generate the response of adaptation and (ii) the response based on adapting to determine multichannel imageOverall response.
Example 2 comprises the purport of example 1, and wherein one or more wave filters are by identical wave filter (identityFilter) composition.
Example 3 comprises any the purport in example 1 and 2, and wherein determines that local difference color vector comprises reallyThe vector of the collinear vector of the filter response calmly and based on each image channel to the definite overall response of each image channel.
Example 4 comprises any the purport in example 1-3, and wherein determines that local difference color vector comprises: reallyDetermine the symmetric form matrix of multichannel image; And the minimum of a value characteristic value of identification and symmetric form matrix or maximum characteristic valueCorresponding characteristic vector.
Example 5 comprises any the purport in example 1-4, and wherein determines that local difference color vector comprises definiteFor the symmetric form of the quadratic form matrix of multichannel image; And minimum of a value characteristic value or the maximum of identification and quadratic form matrixThe characteristic vector that value tag value is corresponding.
Example 6 comprises any the purport in example 1-5, and wherein determines that quadratic form matrix comprises image channelI and j compute matrix, wherein, and wherein qij is at iThe element of matrix A that row and j are listed as, f is the image channel corresponding with the index of f of the filter response based on image channelResponse vector, T is transposed operator, and B is the predefined matrix based on one or more image filters.
Example 7 comprises any the purport in example 1-6, and wherein determines that local difference color vector comprises and make to knowOther characteristic vector standardizes to generate partial error's point color vector.
Example 8 comprises any the purport in example 1-7, and wherein determines each image channel of multichannel imageFilter response comprise to each pixel of multichannel image determine multichannel image each image channel wave filter ringShould.
Example 9 comprises any the purport in example 1-8, and wherein determines each image channel of multichannel imageFilter response comprise that the filter response based on each image channel is paired in the response vector of each image channel next life.
Example 10 comprises any the purport in example 1-9, and wherein to local difference color vector application filteringDevice response comprises the dot product that calculates local difference color vector and filter response.
Example 11 comprises any the purport in example 1-10, and wherein determines that local difference color vector comprises baseDetermine the local difference color vector of normalization in filter response.
Example 12 comprises any the purport in example 1-11, and wherein responds determination module and how logically further suppressThe non-extreme value response in space of the overall response of road image.
Example 13 comprises any the purport in example 1-12, and wherein suppresses the non-extreme value response in space and comprise from manyThe overall response of channel image is removed non-point of interest, and wherein these non-points of interest are identified based on predefined threshold value.
Example 14 comprises any the purport in example 1-13, and further comprises display apparatus module, at meterOn the display of calculation equipment, show the image of instruction overall response.
Example 15 comprises any the purport in example 1-14, and further comprises image capture module, for usingThe filming apparatus of computing equipment is caught captive image, and wherein multichannel image is the image that is hunted down.
Example 16 comprises any the purport in example 1-15, and wherein one or more image filters comprise oneOne or more in order derivative image filter or second dervative image filter.
Example 17 comprises that the method comprises by computing equipment for computing equipment is carried out to the method that multi-channel feature detectsThe filter response of each image channel of definite multichannel image for one or more image filters; By computing equipmentDetermine local difference color vector based on filter response; Responded to local difference color vector filter application by computing equipmentGenerate the response of adaptation; And the overall response of multichannel image is determined in the response based on adapting to by computing equipment.
Example 18 comprises the purport of example 17, and wherein one or more wave filters are made up of identical wave filter.
Example 19 comprises any the purport in example 17 and 18, and wherein determines that local difference color vector comprisesDetermine with filter response based on the each image channel collinear vector to the definite overall response of each image channel toAmount.
Example 20 comprises any the purport in example 17-19, and wherein determines that local difference color vector comprisesDetermine the symmetric form matrix for multichannel image; And minimum of a value characteristic value or the maximum of identification and symmetric form matrixThe characteristic vector that characteristic value is corresponding.
Example 21 comprises any the purport in example 17-20, and wherein determines that local difference color vector comprisesDetermine the symmetric form for the quadratic form matrix of multichannel image; And the minimum of a value characteristic value of identification and quadratic form matrix orThe characteristic vector that maximum characteristic value is corresponding.
Example 22 comprises any the purport in example 17-21, and wherein determines that quadratic form matrix comprises imagePassage i and j compute matrix, wherein, and q whereinijBeThe element of the matrix A of the capable and j of i row, f is that the image corresponding with the index of f of the filter response based on image channel leads toThe response vector in road, T is transposed operator, and B is the predefined matrix based on one or more image filters.
Example 23 comprises any the purport in example 17-22, and wherein determines that local difference color vector comprisesMake the characteristic vector of identification standardize to generate partial error's point color vector.
Example 24 comprises any the purport in example 17-23, and wherein determines each image of multichannel imageThe filter response of passage comprises the filtering of each pixel of multichannel image being determined to each image channel of multichannel imageDevice response.
Example 25 comprises any the purport in example 17-24, and wherein determines each image of multichannel imageThe filter response of passage comprises that the filter response based on each image channel generates response vector to each image channel.
Example 26 comprises any the purport in 17-25, and wherein to local difference color vector filter applicationResponse comprises the dot product that calculates local difference color vector and filter response.
Example 27 comprises any the purport in example 17-26, and wherein determines that local difference color vector comprisesDetermine the local difference color vector of normalization based on filter response.
Example 28 comprises any the purport in example 17-27, and further comprises by computing equipment and suppress to lead to moreThe non-extreme value response in space of the overall response of road image.
Example 29 comprises any the purport in example 17-28, and wherein suppress the non-extreme value response in space comprise fromThe overall response of multichannel image is removed non-point of interest, and wherein these non-points of interest are identified based on predefined threshold value.
Example 30 comprises any the purport in example 17-29, and is further included in the display of computing equipmentThe image of upper demonstration instruction overall response.
Example 31 comprises any the purport in example 17-30, and further comprises by the shooting of computing equipment and fillingPut and catch the image that is hunted down, wherein multichannel image is the image that is hunted down.
Example 32 comprises any the purport in example 17-31, and wherein one or more image filters compriseOne or more in first derivative image filter or second dervative image filter.
Example 33 comprises computing equipment, and it comprises: processor; And memory, it has the multiple instructions that are stored in wherein,These instructions impel computing equipment to carry out any the method in example 17-32 in the time being carried out by processor.
Example 34 comprises one or more machinable mediums, and it comprises multiple instructions stored thereon, theseInstruction causes computing equipment to carry out any the method in example 17-32 in response to being performed.
Example 35 comprises the computing equipment detecting for multi-channel feature, and this computing equipment comprises: for determining for oneThe parts of the filter response of each image channel of the multichannel image of individual or multiple image filters; Be used for based on wave filterThe parts of local difference color vector are determined in response; For responding to generate adaptation to local difference color vector filter applicationThe parts of response; And determine the parts of the overall response of multichannel image for the response based on adapting to.
Example 36 comprises the purport of example 35, and wherein one or more wave filters are made up of identical wave filter.
Example 37 comprises any the purport in example 35 and 36, and wherein for determining local difference color vectorParts comprise for determining with filter response based on each image channel the definite overall response of each image channelThe vectorial parts of collinear vector.
Example 38 comprises any the purport in example 35-37, and wherein for determining local difference color vectorParts comprise: for determining the parts for the symmetric form matrix of multichannel image; And for identification and symmetric formThe parts of the corresponding characteristic vector of the minimum of a value characteristic value of matrix or maximum characteristic value.
Example 39 comprises any the purport in example 35-38, and wherein for determining local difference color vectorParts comprise the parts of symmetric form of the quadratic form matrix for determining multichannel image; With for identification with quadratic form squareThe parts of the minimum of a value characteristic value of battle array or the corresponding characteristic vector of maximum characteristic value.
Example 40 comprises any the purport in example 35-39, and wherein comprises use for the parts of determining quadratic form matrixIn to image channel i and j compute matrixParts, wherein，And wherein qijBe capable at i and the element of the matrix A of j row, f is filter response based on image channel and the index of fThe response vector of corresponding image channel, T is transposed operator, and B is being scheduled to based on one or more image filtersJustice matrix.
Example 41 comprises any the purport in example 35-40, and wherein for determining local difference color vectorParts comprise that the characteristic vector for making identification standardizes to generate the parts of partial error's point color vector.
Example 42 comprises any the purport in example 35-41, and wherein for determining the each of multichannel imageThe parts of the filter response of image channel comprise for the each pixel to multichannel image determines the each of multichannel imageThe parts of the filter response of image channel.
Example 43 comprises any the purport in example 35-42, and wherein for determining the each of multichannel imageThe parts of the filter response of image channel comprise and are paired in each next life for the filter response based on each image channelThe parts of the response vector of image channel.
Example 44 comprises any the purport in example 35-43, and wherein for should to local difference color vectorComprise the parts of the dot product for calculating local difference color vector and filter response with the parts of filter response.
Example 45 comprises any the purport in example 35-44, and wherein for determining local difference color vectorParts comprise for based on filter response determine normalization local difference color vector parts.
Example 46 comprises any the purport in example 35-45, and further comprises for suppressing multichannel imageThe parts of the non-extreme value response in the space of overall response.
Example 47 comprises any the purport in example 35-46, and wherein for suppressing space non-extreme value responseParts comprise the parts for remove non-point of interest from the overall response of multichannel image, and wherein these non-points of interest are based on predefinedThreshold value is identified.
Example 48 comprises any the purport in example 35-47, and further comprises aobvious at computing equipmentShow the parts that show the image of instruction overall response on device.
Example 49 comprises any the purport in example 35-48, and further comprises for the bat with computing equipmentTake the photograph device and catch the parts of captive image, wherein multichannel image is captive image.
Example 50 comprises any the purport in example 35-49, and wherein one or more image filters compriseOne or more in first derivative image filter or second dervative image filter.
Example 51 comprises the computing equipment detecting for multi-channel feature, and this computing equipment comprises: local difference color mouldPiece, determines local difference color vector for the pixel value of the each image channel based on multichannel image; Determine with responseModule, for the pixel value of (i) local difference color vector being applied to each image channel generate adaptation response and(ii) the overall response of multichannel image is determined in the response based on adapting to.
Example 52 comprises the purport of example 51, and wherein determine local difference color vector comprise determine with based on eachThe vector of the collinear vector of the pixel value of image channel to the definite overall response of each image channel.
Example 53 comprises any the purport in example 51 and 52, and wherein determines that local difference color vector comprisesDetermine the symmetric form matrix for multichannel image; And minimum of a value characteristic value or the maximum of identification and symmetric form matrixThe characteristic vector that characteristic value is corresponding.
Example 54 comprises any the purport in example 51-53, and wherein determines that local difference color vector comprisesMake the characteristic vector of identification standardize to generate partial error's point color vector.
Example 55 comprises any the purport in example 51-54, and wherein responds determination module and further suppress manyThe non-extreme value response in space of the overall response of channel image.
Example 56 comprises that the method comprises by computing equipment for computing equipment is carried out to the method that multi-channel feature detectsThe pixel value of the each image channel based on multichannel image is determined local difference color vector; By computing equipment to partial errorThe pixel value that point color vector is applied each image channel generates the response of adaptation; And by computing equipment based on adapt to soundShould determine the overall response of multichannel image.
Example 57 comprises the purport of example 56, and wherein determine local difference color vector comprise determine with based on eachThe vector of the collinear vector of the pixel value of image channel to the definite overall response of each image channel.
Example 58 comprises any the purport in example 56 and 57, and wherein determines that local difference color vector comprisesDetermine the symmetric form matrix for multichannel image; And minimum of a value characteristic value or the maximum of identification and symmetric form matrixThe characteristic vector that characteristic value is corresponding.
Example 59 comprises any the purport in example 56-58, and wherein determines that local difference color vector comprisesMake the characteristic vector of identification standardize to generate partial error's point color vector.
Example 60 comprises any the purport in example 56-59, and further comprises by computing equipment and suppress to lead to moreThe non-extreme value response in space of the overall response of road image.
Example 61 comprises computing equipment, and it comprises: processor; And memory, it has the multiple instructions that are stored in wherein,These instructions impel computing equipment to carry out any the method in example 56-60 in the time being carried out by processor.
Example 62 comprises one or more machinable mediums, and it comprises multiple instructions stored thereon, theseInstruction causes computing equipment to carry out any the method in example 56-60 in response to being performed.
Example 63 comprises the computing equipment detecting for multi-channel feature, and this computing equipment comprises for carrying out example 56-The parts of the method for any in 60.
Claims (25)
1. the computing equipment detecting for multi-channel feature, described computing equipment comprises:
Image filtering module, for what determine for each image channel of the multichannel image of one or more image filtersFilter response;
Local difference color module, for determining local difference color vector based on described filter response; And
Response determination module, generates adaptation for (i) described local difference color vector being applied to described filter responseThe overall response of described multichannel image is determined in response and the (ii) response based on described adaptation.
2. computing equipment as claimed in claim 1, wherein determine described local difference color vector comprise determine with based on oftenThe vector of the collinear vector of the filter response of individual image channel to the definite overall response of each image channel.
3. computing equipment as claimed in claim 1, wherein determine that described local difference color vector comprises:
Determine the symmetric form for the quadratic form matrix of described multichannel image; And
The characteristic vector that identification is corresponding with the minimum of a value characteristic value of described quadratic form matrix or maximum characteristic value.
4. computing equipment as claimed in claim 3, wherein determines that described quadratic form matrix comprises image channel i and j calculatingMatrix, wherein, and
Wherein qijBe capable at i and the element of the matrix A of j row, f is filter response based on described image channel and fThe response vector of the corresponding image channel of index, T is transposed operator, and B is based on described one or more image filteringsThe predefined matrix of device.
5. computing equipment as claimed in claim 3, wherein determines that described local difference color vector comprises the feature that makes identificationVector gauge formats to generate described local difference color vector.
6. computing equipment as claimed in claim 1, the wherein wave filter of each image channel of definite described multichannel imageResponse comprises the filter response of each pixel of described multichannel image being determined to each image channel of multichannel image.
7. computing equipment as claimed in claim 1, the wherein wave filter of each image channel of definite described multichannel imageResponse comprises that the filter response based on each image channel generates the response vector of each image channel.
8. computing equipment as claimed in claim 7, wherein applies described filter response to described local difference color vectorComprise the dot product that calculates described local difference color vector and described filter response.
9. computing equipment as claimed in claim 1, wherein determines that described local difference color vector comprises based on described filteringThe local difference color vector of normalization is determined in device response.
10. computing equipment as claimed in claim 1, wherein said response determination module further suppresses described multichannel imageThe non-extreme value response in the space of overall response.
11. computing equipments as claimed in claim 10, wherein suppress space non-extreme value response and comprise total from multichannel imageNon-point of interest is removed in response, and wherein these non-points of interest are identified based on predefined threshold value.
12. computing equipments as described in any one in claim 1-11, it further comprises display apparatus module, for describedOn the display of computing equipment, show the image of the described overall response of instruction.
13. computing equipments as described in any one in claim 1-11, wherein said one or more image filters compriseOne or more in first derivative image filter or second dervative image filter.
14. 1 kinds of methods for computing equipment execution multi-channel feature is detected, described method comprises:
Determine the filter for each image channel of the multichannel image of one or more image filters by described computing equipmentThe response of ripple device;
Determine local difference color vector by described computing equipment based on described filter response;
Apply described filter response and generate the response of adaptation to described local difference color vector by described computing equipment; WithAnd
The overall response of described multichannel image is determined in response by described computing equipment based on described adaptation.
15. methods as claimed in claim 14, wherein determine described local difference color vector comprise determine with based on eachThe vector of the collinear vector of the filter response of image channel to the definite overall response of each image channel.
16. methods as claimed in claim 14, wherein determine that described local difference color vector comprises:
Determine the symmetric form matrix for described multichannel image; And
The characteristic vector that identification is corresponding with the minimum of a value characteristic value of described symmetric form matrix or maximum characteristic value.
17. methods as claimed in claim 14, wherein determine that described local difference color vector comprises:
Determine the symmetric form for the quadratic form matrix of described multichannel image; And
The characteristic vector that identification is corresponding with the minimum of a value characteristic value of described quadratic form matrix or maximum characteristic value.
18. methods as claimed in claim 17, wherein determine that described quadratic form matrix comprises image channel i and j calculating squareBattle array, wherein, and
Wherein qijBe capable at i and the element of the matrix A of j row, f is filter response based on described image channel and fThe response vector of the corresponding image channel of index, T is transposed operator, and B is based on described one or more image filteringsThe predefined matrix of device.
19. methods as claimed in claim 17, wherein determine described local difference color vector comprise make identification feature toGauge formats to generate described local difference color vector.
20. methods as claimed in claim 14, wherein the filter response bag of each image channel of definite multichannel imageDraw together the filter response of each pixel of described multichannel image being determined to each image channel of described multichannel image.
21. methods as claimed in claim 14, wherein comprise meter to described local difference color vector filter application responseCalculate the dot product of described local difference color vector and described filter response.
22. methods as claimed in claim 14, it further comprises by described computing equipment and suppresses described multichannel imageThe non-extreme value response in space of overall response.
23. one or more machinable mediums, it comprises multiple instructions stored thereon, described instruction is in response to quiltCarry out and cause computing equipment to execute claims the method described in any one in 14-22.
24. 1 kinds of computing equipments that detect for multi-channel feature, described computing equipment comprises:
Local difference color module, determines local difference face for the pixel value of the each image channel based on multichannel imageLook vector; And
Response determination module, generates for the pixel value of (i) described local difference color vector being applied to each image channelThe overall response of described multichannel image is determined in the response adapting to and the (ii) response based on described adaptation.
25. computing equipments as claimed in claim 24, wherein determine described local difference color vector comprise (i) determine forThe symmetric form matrix of described multichannel image and (ii) identification and the minimum of a value characteristic value of described symmetric form matrix orThe large corresponding characteristic vector of value tag value, and
Wherein said response determination module further suppresses the non-extreme value response in space of the overall response of described multichannel image.
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/IB2013/003104 WO2015079282A1 (en) | 2013-11-28 | 2013-11-28 | Method for determining local differentiating color for image feature detectors |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105683996A true CN105683996A (en) | 2016-06-15 |
| CN105683996B CN105683996B (en) | 2019-10-25 |
Family
ID=50434230
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201380080597.9A Active CN105683996B (en) | 2013-11-28 | 2013-11-28 | Method for determining the local differential color of image feature detector |
Country Status (5)
| Country | Link |
|---|---|
| EP (1) | EP3074925A1 (en) |
| JP (1) | JP2016538630A (en) |
| KR (1) | KR101794465B1 (en) |
| CN (1) | CN105683996B (en) |
| WO (1) | WO2015079282A1 (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111931785A (en) * | 2020-06-19 | 2020-11-13 | 国网山西省电力公司吕梁供电公司 | Edge detection method for infrared image target of power equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101510299A (en) * | 2009-03-04 | 2009-08-19 | 上海大学 | Image self-adapting method based on vision significance |
| EP1944731A3 (en) * | 2007-01-12 | 2010-02-24 | Seiko Epson Corporation | Method and apparatus for detecting objects in an image |
| CN102867301A (en) * | 2012-08-29 | 2013-01-09 | 西北工业大学 | Mehtod for getting image salient features according to information entropy |
| WO2013100779A1 (en) * | 2011-12-29 | 2013-07-04 | Intel Corporation | Generalized robust multichannel feature detector |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3496734B2 (en) * | 1994-04-30 | 2004-02-16 | ソニー株式会社 | Edge region detecting apparatus and method |
| US6295373B1 (en) * | 1997-04-04 | 2001-09-25 | Raytheon Company | Polynomial filters for higher order correlation and multi-input information fusion |
| JP3659914B2 (en) * | 2000-10-31 | 2005-06-15 | 松下電器産業株式会社 | Object recognition apparatus, object recognition method, program, and recording medium |
| JP2006285310A (en) * | 2005-03-31 | 2006-10-19 | Kanazawa Univ | Evaluation method of canopy of forest, and its canopy evaluation program |
| JP5465001B2 (en) * | 2009-12-25 | 2014-04-09 | 三菱電機株式会社 | Target estimation device |
| US9165201B2 (en) * | 2011-09-15 | 2015-10-20 | Xerox Corporation | Systems and methods for detecting cell phone usage by a vehicle operator |
-
2013
- 2013-11-28 KR KR1020167010385A patent/KR101794465B1/en active IP Right Grant
- 2013-11-28 JP JP2016526000A patent/JP2016538630A/en active Pending
- 2013-11-28 WO PCT/IB2013/003104 patent/WO2015079282A1/en active Application Filing
- 2013-11-28 CN CN201380080597.9A patent/CN105683996B/en active Active
- 2013-11-28 EP EP13843069.9A patent/EP3074925A1/en not_active Withdrawn
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1944731A3 (en) * | 2007-01-12 | 2010-02-24 | Seiko Epson Corporation | Method and apparatus for detecting objects in an image |
| CN101510299A (en) * | 2009-03-04 | 2009-08-19 | 上海大学 | Image self-adapting method based on vision significance |
| WO2013100779A1 (en) * | 2011-12-29 | 2013-07-04 | Intel Corporation | Generalized robust multichannel feature detector |
| CN102867301A (en) * | 2012-08-29 | 2013-01-09 | 西北工业大学 | Mehtod for getting image salient features according to information entropy |
Non-Patent Citations (4)
| Title |
|---|
| 吴攀超等: "采用局部差分模型描述的彩色图像配准技术", 《西安交通大学学报》 * |
| 曹健: "基于局部特征的图像目标识别技术研究", 《中国博士学位论文全文数据库》 * |
| 李泽辉: "基于图像序列的运动目标检测与跟踪", 《中国优秀硕士学位论文全文数据库》 * |
| 杨婷婷: "视频中运动目标检测与跟踪算法的研究", 《中国优秀硕士学位论文全文数据库》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| EP3074925A1 (en) | 2016-10-05 |
| CN105683996B (en) | 2019-10-25 |
| KR20160060121A (en) | 2016-05-27 |
| JP2016538630A (en) | 2016-12-08 |
| WO2015079282A1 (en) | 2015-06-04 |
| KR101794465B1 (en) | 2017-11-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110852233A (en) | Hand-off steering wheel detection and training method, terminal, device, medium, and system | |
| US11455831B2 (en) | Method and apparatus for face classification | |
| CN105190685B (en) | Method and apparatus for providing the self-adapting data path for being used for computer vision application program | |
| CN110443212B (en) | Positive sample acquisition method, device, equipment and storage medium for target detection | |
| US20140050387A1 (en) | System and Method for Machine Vision Inspection | |
| CN109801590B (en) | Multi-receiving-unit automatic configuration method, device and system for LED display screen | |
| US20170193324A1 (en) | Method and apparatus for identifying salient subimages within a panoramic image | |
| WO2012093381A1 (en) | Camera assembly with an integrated content analyzer | |
| US10062002B2 (en) | Technologies for determining local differentiating color for image feature detectors | |
| JP7429756B2 (en) | Image processing method, device, electronic device, storage medium and computer program | |
| WO2020259416A1 (en) | Image collection control method and apparatus, electronic device, and storage medium | |
| CN105405130A (en) | Cluster-based license image highlight detection method and device | |
| CN110196250A (en) | The detection method and processing equipment of Cr VI based on color component value | |
| CN114842185A (en) | Method, device, equipment and medium for identifying fire | |
| EP2249307A1 (en) | Method for image reframing | |
| WO2021233058A1 (en) | Method for monitoring articles on shop shelf, computer and system | |
| CN114066823A (en) | Method for detecting color block and related product thereof | |
| CN105683996A (en) | Method for determining local differentiating color for image feature detectors | |
| WO2023179520A1 (en) | Imaging method and apparatus, and image sensor, imaging device and electronic device | |
| US20180307910A1 (en) | Evaluation of models generated from objects in video | |
| US20230100268A1 (en) | Quantifying biotic damage on plant leaves, by convolutional neural networks | |
| CN114004809A (en) | Skin image processing method, device, electronic equipment and medium | |
| CN111912349B (en) | Detection method and system for display substrate | |
| CN113592789A (en) | Dim light image identification method, device, equipment and storage medium | |
| CN109934113B (en) | Classroom people counting method and system, device and storage medium thereof |
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 |