# Method, device, computer program, computer system and machine-readable storage device for the representation and retrieval of the object in the image

The invention relates to the representation of the object appearing in the image. Its use in image processing, stored in a multimedia database, allows to provide the technical result in more accurate search objects in images. This technical result is achieved due to the fact that receive a scaled representation of the space of curvature (IPC) for the object outline by the smooth contour of the object, receive at least one additional parameter reflecting the distribution of the shape or mass of the smoothed version of the original curve, and connect the representation to the IPC and the additional parameter as a descriptor of the form object. This optional parameter may correspond to the eccentricity or the roundness of contour, the highest peak in the image in the IPC, it can be based on Fourier descriptors or Zernike moments, etc. 5 C. and 12 C.p. f-crystals, 3 ill. The technical field to which the invention relates the Present invention relates to the representation of an object appearing in a still image or a video image, such as image, stored in a multimedia database, osobennyi In such applications, as a library of images or videos, it is desirable to have effective representation and storage of the contour or shape of objects or parts of objects that appear in still images or video images. A known method based on the shape index and search, uses a scaled representation of the space of curvature (IPC) (CSS). Details view of the IPC can be found in the article "Robust and Efficient Shape Indexing through Curvature Scale Space" (Sustainable and effective index of the form through the space with a curved scale) Proc. British Machine Vision conference, pp. 53-62, Edinburgh, UK, 1996, and "Indexing an Image Database by Shape Content using Curvature Scale Space" (Indexing databases of images via context forms through space with a curved scale) of the OEWG. IEE Colloquium on Intelligent Databases, London 1996, both written F. Mokhtarian, S. Abbasi, and J. Kittler, bibliographic details are provided here as a reference.View the IPC uses a function of curvature for the path of the object, starting from an arbitrary point on the contour. This function of the curvature is studied as the shape of the path unfolds through a series of deformations, which smooth out the shape. Specifically, we calculate the zero crossing for the derivative of the function cigratte, known as the space is curved scale, where the x-axis represents the normalized arc length of the curve and the y axis is the expansion parameter, specifically, by setting the applied filter. The points on this graph form a loop response of the circuit. Each of the convex or concave part of the contour of the object corresponds to the loop in the image of the IPC. The coordinates of the most prominent peaks of the loops in the image of the IPC is used as a representation of the contour.To search for objects stored in the database image that is consistent with the shape of the input object, we compute the representation of the IPC input object. The similarity between the input form and memorized forms is determined by comparing the position and the height of the peaks in the respective images of the IPC using the algorithm mates.From the first of the above-mentioned article it is also known to use two additional parameters - roundness and eccentricity of the original form is to be excluded from the pairing process forms with significantly different parameters of roundness and eccentricity.The problem with the view described above is that the search accuracy can sometimes be low, especially for curves to Puglia curves.The object of the present invention is to introduce an additional tool description form for "contour shape of the prototype". This shape of the prototype is defined here as: 1) the original form, if the circuit has no convexity or concavity (i.e., for example, in the image of the IPC no peaks), or 2) the shape contour after smoothing is equivalent to the highest peak in the image of the IPC.Note that the shape of the prototype is always convex.For example, the shape of the prototype can be described by invariants, based on the moments of the area, as described in the article "Visual Pattern Recognition by Moments Invariants" (Visual pattern recognition by moment invariants, IEEE Transactions on Information Theory, Vol. IT-8, 179-187, 1962, written by M. K. No, the bibliographic data of which are given here as a reference, or by using Fourier descriptors, as described in the article "On Image Analysis by the Methods of Moments" On image analysis by the methods of moments, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 10, No. 4, July 1988, written Cho-Huak The bibliographic data of which are given here as a reference, or by using parameters such as eccentricity, roundness, etc., In the above-mentioned known methods eccentricity and army circuit prototype", which is different for curves having at least one peak of the IPC. Another difference is that in the known method, the eccentricity and circularity are used to exclude some forms of mates for finding similarity, as here, the applicant uses them (in addition to the peaks of the IPC) to get the value of the similarity measure. Finally, the applicant extends the additional parameters used in the pairing process, to the moment invariants, Fourier descriptors and Zernike moments.As a result of implementation of the invention can be more accurate.The invention is a Method of representing an object appearing in a still image or video image, by processing corresponding to the image signals, according to one object of the present invention, includes receiving a representation of the space with a curved scale (IPC) for the contour of the object by smoothing the contour of the object, receiving at least one additional parameter reflecting the distribution of the shape or mass of the smoothed version of the original curve, and link representation of the IPC and the additional parameter as a descriptor of the form object.In the proposed method of attachment is ornately parameter may relate to a flattened contour, corresponding to the highest peak in the image of the IPC.An additional parameter may correspond to the eccentricity of the path.An additional parameter may correspond to the circularity of the contour.At least one additional parameter can be used based on the view pane.An additional parameter may be a torque invariant region.An additional parameter may be based on Fourier descriptors.An additional parameter may be based on Zernike moments for the area covered by the path.Way to represent the set of objects appearing in a still image or video image, by processing signals corresponding to images, according to the second object of the invention contains, for each object, determining whether there is a significant change in the curvature in the contour of the object, and if the curvature in the contour of the object there are significant changes, then the descriptor form, and if the curvature in the contour of the object there are no significant changes, then the descriptor of the form, including at least said additional parameter reflecting the shape of the contour of the object.acitelli changes in curvature, can be based on moment invariants areas, the Fourier descriptors or the Zernike moments of the contour.Way of the search object in a still image or video image, by processing signals corresponding to images, according to the third object of the invention contains an introduction request in the form of a two-dimensional path, the receive descriptor mentioned circuit, the comparison of the mentioned query descriptor with each descriptor for a stored object using pairing, use the values of the IPC and more options to obtain the similarity measure, and selecting and displaying at least one result corresponding to the image containing the object for which the comparison indicates a degree of similarity between the query and the above-mentioned object.According to the method according to the third object of the invention to measure the similarity can be based on M, where M=a*GP-S+CSS S, where GP-S is a measure of similarity between the additional parameters of the contours of the compared object, CSS-S - a measure of similarity between the IPC values for the contours of the objects being compared, and is a constant.According to the method according to the third object of the present invention and may depend on the number and height of the peaks of the IPC.As can be equal to 1, when the IPC.Way of the search object in a still image or video image, by processing signals corresponding to images according to a fourth object of the invention contains the calculation of the measure of similarity between the contours of two objects using a view IPC-mentioned circuits and additional parameters reflecting the distribution shapes or masses in its original contour and a smoothed version of this circuit.A brief description of the drawings Fig.1 is a block diagram of the system base video.Fig.2 is a drawing of the outline of the object.Fig.3 is a representation of the IPC circuit according to Fig.2.The best implementation of the invention, the First implementation of Fig. 1 shows an automated system base video data according to this invention. This system includes a control unit 2 in the form of a computer, the display unit 4 in the form of a monitor, a pointing device 6 in the form of a mouse, the base 8 of the image data including the stored still images and video images, and a database 10 data descriptor, storing the descriptors of the objects or portions of objects appearing in the images stored in the database 8 of the image data.The descriptor for each caller online who are in the 10 data descriptors. The control unit 2 receives the descriptors in the process under the control of appropriate software embodying the method described below.First for loop of the given object to have a clear understanding of the IPC path. This is done by using a known method as described in one of the aforementioned articles.More specifically, the circuit is expressed by the representation={(x(u) y(u), u[0, 1]}, where u is the parameter of the normalized arc length.Contour smoothed by convolutionwith kernel g (u,) The Gaussian identifier. Zero crossing of the curvature of the unfolding curve check as. Zero crossing is identified using the following expressions for the curvature:,

where X(u,)=x(u)*g(u,); Y(u,)=y(u)*g(u,);

X

_{u}(u,)=x(u)*g

_{u}(u,); X

_{uu}(u,)=x(u)*g

_{uu}(u,).In the above expressions, * represents convolution, and subscripts represent derivatives.The number parasocial aboveis convex curve without crossing zero.Point (u,) of zero-crossings based on the graph, known as the image space of the IPC. This is reflected in many of the characteristics curves of the original circuit. Peaks characteristic curves are identified and the corresponding coordinates stand out and be remembered. In General, this gives a set of n coordinate pairs [(x1,y1), (x2,y2),..., (xn,yn)], where n is the number of peaks, xi is the position of the arc length of the i-th peak, yi is the height of this peak. These coordinates peaks represent the view of the IPC.In addition to presenting the IPC, additional parameters are associated with this form to obtain a descriptor of the form. In this implementation of the additional parameters are the eccentricity and circularity "field prototype" for this form, where the scope of the prototype of this form represents the outline of this shape after the final step of smoothing, i.e. at the point equivalent to the value ofthe highest peak. For the scope of the prototype can be selected and other values of. The result is a descriptor of the form to form S in the form: {EPR, CPR, PEAK}, where EPR is the eccentricity of the field prototype, CPR - rounded is accordance with this invention.Here, the base 10 data descriptors in the system of Fig.1 stores the descriptors form obtained according to the above method.The user initiates the search by drawing the contour of the object on the display using a pointing device. The control unit 2 then obtains a handle to the form of the input circuit as described above. The control unit then performs interfacing comparison with each descriptor form, stored in this database.Assume that the input circuit of the form S1 is compared with the memorized shape S2, and S1 and S2 are the corresponding descriptor:

S1: {EPR1, CPR1, REAK},

S2: {EPR2, CPR2, REAK},

where EPR means the eccentricity of the field prototype, CPR means the roundness of the field prototype, a PEAK mean the set of coordinates of the peaks in the image of the IPC (this set may be empty). The measure of similarity between two shapes is computed as follows.M=a*(abs(EPR2-EPR1)/(EPR2+EPR1))+b*abs((CPR2-CPR1)/((CPR2+CPR1))+SM(PEAKS1, PEAKS2),

where a and b are two coefficients, S - a standard measure of similarity defined on two sets of peaks [1], a abs denotes the absolute value. SM is calculated by using a known algorithm mates which you can use the algorithms described in the aforementioned article the curve

_{i}image and model curve

_{m}and their respective sets of peaks {(xi1,yi1), (xi2, yi2), . . ., (xin,yin)} and {(xm1,ym1) and (xm2, ym2),..., (xmn,ymn)} is computed measure of similarity. This measure of similarity is defined as the total price of a pair of peaks in the model with peaks in the image. The pair which minimizes the total price is determined using dynamic programming. The algorithm recursively matches the peaks from the model with the peaks of the image and calculates the price of each such pair. Each peak model can be paired with only one peak of the image, and each peak of the image can be paired with only one peak model. Some of the peaks of the model and the image may remain unpaired, and, for each unpaired peak is assigned a penalty price. Two peaks can be paired if their horizontal distance less than 0.2. The price of the pair represents the length of a straight line between two adjacent peaks. Price unpaired peak is its height.More details on this algorithm works by creating and expanding a tree structure where the nodes correspond to the paired peaks:

1. To create the initial node, consisting of the most high and the falls in the 80% of the maximum of the maximum peaks of the image, create additional home site.3. To initialize the initial price of each node created in 1 and 2, to the absolute difference of y-coordinates of the peaks of the model and the image associated with this node.4. For each starting node 3 to compute the alpha shift IPC, defined as the difference (horizontal) coordinates x peaks model and image, paired in this initial node. The shift parameter will be different for each node.5. For each starting node to create a list of peaks model and the peaks of the image. This list contains information which peaks still be mate. For each starting node to mark the peaks associated to this node as "conjugate", and all other peaks as "unpaired".6. Recursively expand the node of the lower price (starting from each node created in steps 1-6 and accompany their child nodes) until then, until the conditions of paragraph 8. To expand a node use the following procedure:

7. The expansion of the node.If there are unpaired at least one peak of the image and one peak model: select the maximum IPC curve image on the greatest scale, which does not involve (xip, yip). To make the add option is now selected peak has coordinates (xip-alpha, yip). To determine the position of the nearest peak of the curve model, which does not involve (xms, yms). If the horizontal distance between these two peaks is less than 0.2 (i.e., |xip-alpha-xms|<0,2), to produce the pairing of these two peaks and to determine the price of this pair as the length of a straight line between these two peaks. Add the price of this pair to the total price of this node. Delete the paired peaks from the appropriate lists, marking them as "connected". If the horizontal distance between these two peaks is greater than 0.2, this peak (xip,yip) image can not be paired. In this case, add its height yip to the total price and to delete only this peak (xip,yip) from the list of peaks in the image by marking it as "unpaired".Otherwise, there are only the peaks of the image or there are only peaks model, the remaining unpaired):

To determine the price of a pair as the height of the highest unpaired peak image or model and remove this peak from the list.8. If after expanding a node in p. 7 in both lists images and models no non-paired sites, the pairing is completed. The price of this node is a measure of similarity between the curve of the image and the model. Otherwise, go to p. 7 and expand the node N. the curve model. The final value of the pair is the smallest of the two.The above steps are repeated for each model in the database.Similarity measure appearing in the result of the comparison pair, are arranged in order, and the objects corresponding to the descriptors having a similarity measure that indicates the very close pair (i.e., here the lowest similarity measure), then appear on the display unit 4 to the user. The number of objects to be displayed, can be set in advance or selected by the user.In an alternative embodiment to describe a form field prototype" can be used in a variety of settings. For example, you can use three Fourier coefficient for the curve. The measure of similarity may be determined as follows:

M=a*EUC(F1,F2)+SM(PEAKS1+PEAKS2),

where EUC is the Euclidean distance between the vectors F1 and F2, is generated from three main Fourier coefficients for the form image and the model, and is a constant, SM represents the degree of similarity to the peaks of the IPC, calculated using the method essentially described above.Industrial applicability

The system according to this invention can, for example, be provided in the image library. Alternatively, the database may buy as a telephone line, or a network such as the Internet. Database of images and descriptors may include, for example, in a permanent storage device or data on a portable storage medium such as CD-ROM or DVD.Components of the system as described can be implemented in software or hardware form. Although the invention is described in the form of a computer system, it may be embodied in other forms, for example using a specialized IP.Concrete examples of ways of representing a two-dimensional shape of the object and the methods of calculation of values representing the similarity between the two forms, but can be any suitable such methods.The invention can also be used, for example, for a pair of images of objects for the purposes of verification or for filtering.

Claims

**Same patents:**

FIELD: computer science, possible use for measuring coordinates of light objects and processing their movement trajectory.

SUBSTANCE: device containing transmitting television tube, blocks forming spiral scanning of television tube ray, and signals processing block, additionally has clock pulse counter and video pulse counter, and also analog-digital converter and visual observation block, constructed on basis of piezo-deflectors with optical reflectors at the end. It is possible to output information about 72000 light objects at 100 Hz frequency.

EFFECT: possible measurement of polar coordinates of light objects, simultaneously located within objective observation field, their brightness, transverse size of image and distance from axis within observation field of each frame (order number).

3 dwg, 1 tbl

FIELD: physics, processing of images.

SUBSTANCE: invention is related to methods of television image processing, namely, to methods of detection and smoothing of stepped edges on image. Method consists in the fact that pixels intensity values (PIV) of image are recorded in memory; for every line: PIV of the current line is extracted; PIV of line that follows the current line is extracted; dependence of pixel intensity difference module dependence (PIDMD) is calculated for the mentioned lines that correspond to single column; PIDMD is processed with threshold function for prevention of noise; "hill" areas are determined in PIDMD; single steps are defined out of "hill" areas; PIV of line that is next nearest to the current line is extracted; for current line and line next nearest to the current line operations of "hill" areas definition are repeated; for every part of image line that is defined as single step, availability of stepped area is checked in image in higher line, if so, these two stepped areas are defined as double stepped area (DSA); parts of DSA lines are shifted in respect to each other, and DSA is divided into two single steps; values of line pixels intensity are extracted for the line that is located in two lines from the current line, and operations of "hill" areas definition are repeated; single steps are smoothened by averaging of pixel intensity values.

EFFECT: improvement of quality of image stepped edges correction.

2 dwg

FIELD: information technology.

SUBSTANCE: snap-shot of a graphic image is reduced by resising by 4 times. The snap-shot is compressed and stored as a compressed file, which can be decompressed and increased by 4 times. The initial snap-shot of the graphic image is then superimposed onto a decompressed increased snap-shot. Differences of pixel values between the initial snap-shot of the graphical image and the decompressed increased snap-shot are searched for, from a given contrast, based on contrast elements using arithmetical subtraction. After that a snap-shot with contour values is obtained, and contrast elements are compressed and stored as a compressed file.

EFFECT: increased efficiency of compressing files and reducing amount of memory required with retention of definition of the graphical image after decompression of files.

8 dwg

FIELD: information technology.

SUBSTANCE: present invention relates to digital processing of images and can be used in photographic, video, optical-location and optical-electronic engineering for identifying images from their contours on digital images. The method of noise-immune gradient detection of contours of objects on digital half-tone images is based on preliminary assessment of the location of impulse noise on the image, after which four auxiliary masks are formed, as well as four corresponding control vectors. Using vector data, coefficients of the corresponding four differently aligned Previtt masks are changed. After this the approximate value of the module of the image gradient is calculated using mask data, and contours of the objects on the image are obtained from its threshold conversion.

EFFECT: increased stability of gradient operators for contour detection to impulse noise, arising from many conditions for transferring and converting images.

6 dwg