Method and apparatus for motion estimation

MYPI1996005562A0Inactive Publication Date: 1997-07-11SAMSUNG ELECTRONICS CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
MYPI1996005562
Authority / Receiving Office
MY · MY
Patent Type
Applications
Current Assignee / Owner
Priority Date
1996-01-11
Filing Date
1996-12-31
Publication Date
1997-07-11
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Conventional motion estimation methods, especially those using integer-pixel units, introduce errors during motion compensation, and are inefficient in processing data for finer resolution scales, leading to prolonged processing times and reduced accuracy.

Method used

A method and apparatus that utilize error values generated by comparing image segments between frames to estimate motion vectors at a finer scale of resolution, specifically using a combination of integer-pixel and half-pixel units to interpolate and sum motion vectors, thereby improving data compression efficiency.

Benefits of technology

This approach enables more accurate and efficient motion estimation and compensation by reducing data processing time and enhancing the precision of motion vector detection at finer resolution scales, leading to improved data compression and transmission efficiency.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

A METHOD FOR MOTION ESTIMATION INCLUDES STEPS FOR CONSTRUCTING A FIRST IMAGE SEGMENT REPRESENTING A PORTION OF PRESENT IMAGE FRAME AND A SECOND IMAGE SEGMENT HAVING A PLURALITY OF IMAGE SEGMENTS OF AN ADJACENT FRAME, COMPARING SAID PLURALITY OF IMAGE SEGMENTS OF SAID SECOND IMAGE SEGMENT WITH THE IMAGE SIGNAL WITHIN THE FIRST IMAGE SEGMENT, AND GENERATING A NUMBER OF ERROR VALUES, DETECTING A FIRST MOTION VECTOR AT ONE SCALE OF RESOLUTION USING THE GENERATED ERROR VALUES, OUTPUTTING THE ERROR VALUE ASSOCIATED WITH A REFERENCE SEGMENT POINTED BY THE FIRST MOTION VECTOR, AND THE ERROR VALUES CORRESPONDING TO THE REFERENCE SEGMENT SHIFTED BY ONE UNIT OF SELECTED RESOLUTION RESPECTIVELY IN SELECTED DIRECTION(S), GENERATING A SECOND MOTION VECTOR IN FINER SCALE OF RESOLUTION USING THE OUTPUT ERROR VALUES, AND SUMMING THE FIRST AND SECOND MOTION VECTORS. FIGURE 3B
Need to check novelty before this filing date? Find Prior Art

Description

METHOD AND APPARATUS FOR MOTION ESTIMATIONThe present invention relates to a method and apparatus for performing motion estimation, and more particularly, though not exclusively relates to a method and apparatus for extremely detailed motion estimation which, for an estimation of motion vector(s) at a scale of finer resolution, uses error values generated at an estimation of motion for determining a motion vector, at a scale of lower resolution especially differential pulse code modulation (DPCM) method for coding a digital image signal.Generally, there are several coding methods for iently compressing transmission data in an processing apparatus using subsequent digital , such as high definition television (HDTV), (digital video tape recorder), and multiThe DPCM method codes the digital image data correlations which are present between of the image signal.further effid image signal image signals Digital VTR media, etc. signal using adj acent framesIn the case of coding a difference signal between adjacent frames representing different times using the DPCM method, it is disadvantageous that much more data needs to be coded for an active image rather than a still image. However, by recognising a movement location from a certain segment of an adjacent frame to a specific segment in the present frame, that is, the best similarity between a specific image of the present frame and a certain imacje of the adjacent frame, efficient data compression can be performed because signal difference values of similar segments between frames can be minimised after motion compensated compensation. Accordingly, in the motionDPCM method, difference data between a predetermined corresponding coded, there to be enhanc vector repre image segmen the present an adjacent iframe.image segment of the present frame and the image segment of the adjacent frame is □y permitting efficiency of the transmission In the motion estimation method, a motion sents a moving direction and magnitude of the t obtained by comparing an image segment of frame with the corresponding image segment ofThere are several methods for generating motion estimates ir terms of a pixel unit. Generally, since en two frames separated in time may be small with respect: to an integer-pixel unit, errors may be introduced during motion compensation when using the motion vector of an integer-pixel unit. Therefore, a r of a sub-pixel unit is used for reducing nsation errors. A half-pixel unit, as one estimation method.of sub-pixel units, is commonly used for the motion estimation. A motion estimation method using unit in accordance with a conventional method fly described.motion vectc motion compe a half-pixel will be brie aratus of Figure 1 comprises an image segment 1 for receiving an image signal of the present onstructing an image segment composed of a f pixels; search range constructor 2 for ie image signal of an adjacent frame and the image segment composed of a plurality of irst search unit 3 for receiving the image ut from the image segment constructor 1 and range constructor 2 and detecting a first tor MV1 of a pixel unit; a half-pixel g unit 4 for receiving image signals output rch range constructor 2 and the first search dalculating image signal values in half-pixelThe app constructor frame and c plurality o receiving t constructing pixels; a f signals outp the search motion vec interpolatin' from the se unit 3 and locations around reference pixels, using the image segment, which within the adja will be referred to as a reference segment, ι tent frame which is determined by the first motion vector MV1 detected from the first search unit 3 and its surrounding pixels; a second search unit 5 for receiving the image signals output from the image segment constructor 1 and the half-pixel interpolating unit 4 and outputting a second motion vector MV2 of a half-pixel unit; and a summing unit 6 for receiving and summing the output signals MV1, MV2 from the first search unit 3 and the second search unit 5, respectively, and outputting a motion vector MV.The image segment constructor 1 receives the image signal of the present frame and stores the input signal in segments. Simultaneous with inputting the image signal of the present frame, the search range constructor 2 receives the image signal from an adjacent frame and constructs a search range segment. The first search unit 3 searches for a segment within the search range segment which best matches the secjment in the present form. By comparing data in a pixel unit between two segments the first motion vector MV1 of an integer-pixel unit generated from the first search ur. the half-pixel interpolating it 3 is output to the summing unit 6 and interpolating unit 4. The half-pixel unit 4 linearly interpolates out the reference pixels of a designated segment within the adjacent frame using the first motion vector MV1 of an integer-pixel unit and the surrounding pixels, and then outputs the calculated values to the second search unit 5.above conventional system, the image segment I2block type composed of Nj X N2pixels, he search range may be composed of Μ, X M2In the may be a Nj XAdditionally, til pixels. Alterriately, the image segments and the search range may be constructed in accordance with the shape and size of the objects within the image. Furthermore, in the case of a uni-directional prediction which requires one adjacent frame for reference as in P-frame in MPEG or in B-frame that involves only backward prediction, the image segment of the present frame is compared with an image segment within the search range constructed either within the previous or future frame. However the motion estimation previously mentioned is also applicable to the bi-directional prediction which involves both previous and future frames. In the case of a bi-directional predictions as in B-frame in MPEG, the image segment of the present frame is compared with an image segment within the search ranges constructed within the previous frame and the future frame.Another image segment accordance wi constructed im be of any size conventional method for constructing the and the search range is accomplished in th the objects within the image. Thus, age segment and the search range segment may and shape .In Figure x indicates 5 moves the corresponding2, 0 illustrates pixel unit intervals and half-pixel unit intervals. The second unit image segment of an adjacent frame to the first motion vector MV1 of an integer-pixel unit to every direction in half-pixel units, respectively, motion compen searches for a location having a minimum sation error among locations x for the obtained eight: half-pixel motion vectors and the first motion vector MV1, and outputs one of the vector components (-1 / 2, 0, 1 / 2), that is a minutely controlled value in a half-pixel unit. The summing unit 6 sums the first motion vector MV1 in integer-pixel units output from the first search unit 3 and the second motion vector MV2 units from the second search unit 5, to ate a complete motion vector MV for the ion. For example, the summing unit 6 sums component (-1 / 2) of the second motion a half-pixel unit from the second search motion vector (MV1=(x, y)) of a pixel unit ; search unit 3, and then determines the (MV=(x-l / 2, y)).in half-pixel thereby gener motion estimat the horizontal vector MV2 of unit 5 with thi from the firs motion vectorIn the which involves same procedure predictions .ase of bi-directional motion-estimation, previous and future adjacent frames, the is applied to both of forward and backward interval unit, pixel values aThe above-described conventional method determines the location of segment using the motion vector of a pixel linearly interpolates but corresponding t half-pixel locations using the determined segment and surrounding pixels of the segment. The method uses the interpolated half-pixel values and the segment pixel values of the present frame determined by the motion vector in integer-pixel units, and selects the most similar segment among eight half-pixel motion positions, of Figure 2. Accordingly, the conventional drawback in that it takes too long in a for estimating the motion vector in halfthat is, x method has a processing dat pixel units. This drawback is not confined to the case of motion vector estimation at half-pixel unit, but is a general problem for a motion vector estimation at a resolution finer than a pixel unit.An aim invention is which makes e motion possibl bf preferred embodiments of the present to provide a method for motion estimation fficient compensation of extremely detailed e, wherein error values at one scale of e used for estimation of motion vectors at a scale of resolution, comparing error values s a result of comparing data between image segment within the present frame and within the search range to detect the motion el unit, and estimating the motion vector at ion.resolution ar more detailed generated a predetermined image segment vector in pix finer resolutAnother aim of preferred embodiments of the present invention is to provide an apparatus for motion estimation making efficient compensation of extremely detailed motion possible, by employing error values used for estimation of s between frames, by generating error values to one-scale of resolution by comparing data between predetermined image segment within the present image segments with the adjacent frame(s), comparing generated errors, and estimating the motion vector at a finer scale motion vector correspondingAccordin there is prov between adjac image, the me g to ided ent thod the present invention in a first aspect, a method for comparing an image signal frames and estimating a motion of the comprising the steps of:construe portion of pr having a plur frame with re image segment future image f rame;ting a first image segment representing a asent image frame and a second image segment ality of image segments of a previous image spect to the present image frame and a third having a plurality of image segments of a frame with respect of the present image g said plurality of image segments of said segment with the image signal within the segment, and simultaneously comparing said comparin second image first image plurality of image segments of said third image segment signal within the first image segments, and umber of error values;first motion vector at one scale of with the image generating a n detecting resolution using the generated error values ,g the error value associated with a ent pointed by the first motion vector, and es corresponding to the reference segment unit of selected resolution respectively in :ion (s) ,· output t ih1reference segmi the error valt. shifted by one selected direc generating a second motion vector in a finer scale of resolution using the output error values; and summing tne first and second motion vectors.Suitably, said generating step further comprises the sub-steps of:detecting a component of said second motion vector in a first direction using the error value of said reference and the error values of said surrounding which are positioned in said first respect to a centre of said reference image image segment image segment^ direction with segment; and detecting a component of said section motion vector in a second direction using the error value of said reference image segment and the error values of said surrounding image segments which are positioned in said second direction with respect to the centre of said reference image segment.Suitably, direction are respectively.said first direction and said second horizontal direction and vertical directionSuitably in said first motion vecto: location of t error value., said component of said second motion vector direction and said component of said second r in said second direction indicate the he image segment corresponding to a minimumSuitably previous fram said previous with respect to frame is an immediately said present frame.Suitably plurality of , yxv * ___________ _________ frames in time from said present frameSuitably, frame with re said future frame is an immediately future ^pec.t to said present frame.Suitably, plurality of f said 'rames future frame is separated in time from said present frame.bySuitably, each of said plurality of image segments of said second image segment has a segment size equal to said first image signal.Suitably, said one scale of resolution is a one pixel-unit and said finer scale of resolution is a half pixel-unit.According aspect, there image signal motion of the to the present invention in a second is provided an apparatus for comparing an between adjacent frames and estimating a image, the apparatus comprising:receiving the image signal of a present Structing a first image segment;means for frame and con means for receiving the image signal of two adjacent frames for each received image signal corresponding to an adjacent frame, constructing a second image segment from a previous image frame with respect to the present image frame and a third image segment from a future image frame with respect to the present image frame;search means for receiving the image signal output respectively from the first image segment constructor and the second image segment constructor and the third image segment constructor, detecting a first motion vector at one selected scale of resolution units, determining a reference segment within said second image segment and within said third image segment in accordance with said first motion vector, and simultaneously for said second and third image segments comparing the image signals of surrounding image segments, which are generated by moving image segment in selected direction(s) by the selected scale respectively, and the reference ima^e segment with the image signal of the first the reference one unit of image, and out putting error values obtained as a result of respective comparisons;eans for receiving the errors output from ns and generating a second motion vector at of resolution than the selected scale; and control rti< the search mea a finer scale summing vectors, for eans receiving the first and second motion c.dding and outputting them.Suitably, said control means further comprises:controller for comparing said error value of image segment with the error values of said age segments which are positioned in a first respect to a centre of said reference image btecting a vector component in said first said second motion vector; and a first said reference surrounding im direction with segment and d direction of a second controller for comparing said error value of said reference image segment with the respective error values of said surrounding image segments which are positioned in a second direction with respect to the centre of said reference image segment and detecting a vector component in said second direction of said second motion vector.Suitably, direction are respectively.Suitably, a first corresponding error values segments which image segment;said first direction and said second horizontal direction and vertical direction said controller further comprises:Subtracter for subtracting the error value to said reference image segment from the corresponding to said surrounding image are located to one side of said reference a second corresponding error values segments whic reference imac subtracter for subtracting the error value to said reference image segment from the corresponding to said surrounding image a are located to the other side of said e segment;a first multiplier for multiplying a respective first output signal of said second subtracter by a predetermined coefficient, and producing a first multiplied value;a second second output predetermined multiplied value multiplier for multiplying a respective signal of said first subtracter by said coefficient, and producing a second a first signal and sa comparator, receiving said second output id first multiplied value, for comparing respective magnitudes of the received first comparator inputs, thereby providing a first binary signal as a comparison result;a second signal and said respective mac inputs, thereb comparison res comparator, receiving said first output second multiplied value, for comparing the nitudes of the received second comparator y providing a second binary signal as a ult; and e, receiving said first and second binary said first and second comparators, for performing a logical NOR function, ing an operated value indicative of said a NOR gat signals from respectively, thereby produc second motion vector.said previous frame is an immediately with respect to said present frame.said previous frame is separated by a rames in time from said present frame.said future frame is an immediately future oect to said present frame.Suitably, previous frameSuitably, plurality of fSuitably, frame with resSuitably said future frame is separated by a plurality of frames in time from said present frame.said second image segment has a plurality dnt s.Suitably of image segmiSuitably] each of said plurality of image segments of said second image segment has a segment size equal to said first image segment.Suitably pixel-unit anci pixel-unit.said one scale of resolution is a one said finer scale of resolution is a halfPreferred described, by drawings wherd 3B and 4 of t>Figure ] conventional estimation.Figure 2 unit and a hajFigure apparatus for according to appendant cla.Figure apparatus for with a prefer embodiments of the present invention are way of example only, with reference to the in with reference to and as shown in Figures e accompanying drawings :is a block diagram for illustrating a apparatus used for half-pixel motion is a conceptual diagram for showing a pixel f-pixel unit.$A is a block diagram illustrating an performing motion estimation that while not the present invention as defined in the ^ms is useful for an understanding thereof.3B is a block diagram illustrating an performing motion estimation in accordance red embodiment of the present invention.Figure 4 horizontal ha(If is a detailed block diagram illustrating a pixel controller of Figures 3A-3B.Figures 5(a) and 5(b) are conceptual diagrams showing a method for determining motion vectors of a half-pixel unit in the horizontal half-pixel controller.on is now made for a case in which a first selected at any scale and a second motion iner scale are selected to be pixel unit half-pixel unit respectively. This example e applied to any other selected scales.Descripti motion vector vector at a f resolution and however, can bA motion present frame frame is repre vector of a predetermined segment In a corresponding to a segment in an adjacent Rented as MV=(x,y). Here, the motion vectorMV has a horizontal component of motion vector x and a vertical component of motion vector y.A is a block diagram illustrating an notion estimation.Figure 3 apparatus forThe apparatus of Figure 3A comprises an image segment constructor 3'. for receiving an image signal of the present frame and constructing image segments composed of a plurality of pixels; search range constructor 32 for image signal of an adjacent frame and Lmage segments composed of a plurality of unit 33 for receiving the image signals :he image segment constructor 31 and the receiving the constructing i pixels; search output from t search range constructor 32, for detecting a first motion pixel units by comparing image signals segments within the segment constructed by ge constructor 32 and the image segment in frame, and outputting error values vector MVl in between image the search ran the present corresponding to an image segment pointed by the detected MV1, which will be called as a reference segment and is generated by moving the reference segment in horizontal and vertical directions by one pixel; half-pixel control means 34 for receiving the errors output from the search unit 33 and calculating a second motion vector MV2 in respective palf-pixel horizontally; motion vector motion vector interval vertically and and summing unit 3 7 receiving the first MV1 from the search unit 33 and the second MV2 from the half-pixel control means 34, summing the respective vectors and outputting a resultant motion vector MV.The halfhalf-pixel co component of half-pixel cc component of pixel control means 34 includes a horizontal ntroller 35 for detecting a horizontal tthe second motion vector MV2 and vertical ntroller 36 for detecting a vertical second motion vector MV2.the image signals for two adjacent frames time, the image segment constructor 31 taage signal of the present frame and stores al in a first segment. Simultaneously with image signal of the present frame, the onstructor 32 receives the image signal of me and stores it in a second segment, which may be larger than the first segment. The receives respectively the image signals of bond segment and the first segment output range constructor 32 and the image segment respectively. In the case of a P-frame, is made between the present frame and a however in the case of a B-frame the made between the present frame, and a future ion to the comparison between the presentOf the separated in receives the i the input sign inputting the search range c an adjacent fra.' advantageously search unit 33 the stored se from the search constructor 31 the comparison previous frame comparison is frame in addit, frame and a previous frame. Thus, in the case of a Bframe, the search range constructor 32 constructs a second previous frame and a third segment of a Accordingly, the comparison between the and the previous frame and then between and future frame or vice versa or the comparisons may occur simultaneously as shown in Figure 3B.segment of a future frame. present frame present frameThe following is a description in which the first segment and the second segment have the form Ν, X N2pixels and Μι X M2respectively. Furthermore, for simplicity of explanation, an integer-pixel unit is selected as one scale of resolution and a half-pixel unit is selected as the finer scale of resolution.A motion estimation error obtained by estimating the motion vector Po, Motion esti in horizontal respectively, respectively P!□f an integer-pixel unit is represented by mation errors which are generated by moving and vertical direction by one pixel, on the basis of the estimated vector, are P.,' , P_! and Pj. Mean absolute error MAE, or mean squared error MSE advantageously can be used for calculating motion estimation errors. Equations which preferably can be used to calculate the horizontal motion estimation errors (Po, P,, P^) by the MAE are as follows:N-XN-X*.-ΣΛ1“0 £2 \Y(Nk+m,Nl+n) -Y'(Nk+m+χ, Nl+n+y) . . . (1)N-l Nj- _ \Y(Nk+zn,Nl+n) -Υ!(Nk+m+x+1, Nl+n+y) jn-o η... (2)Ν-1 Ν-1Ρ_1=^2 £ \ Y(Nk+m,Nl+n) -γ'(Nk+m+x-1,Nl+n+y) m-o ή»ο . (3)Where Υ frame, Υ' rep Ν]=Ν2=Ν represe (k, 1) represe the frames, motion estimat the same method represents the image signal of the present resents that of the adjacent frame, and nt segment size. In the above equations, nts position from the top left position of It will be appreciated that the vertical ion errors (P / , P. / ) can be calculated by as noted by the above listed equation.The search unit 33 compares, in pixel units, the data magnitude between the image segment of the present frame and a plurality of adjacent image segments, which are the search range of the adjacent frame and sizes, and calculates motion estimation suit of respective comparisons. The first motion vector NV1, in integer-pixel units is determined as a segment location where the motion estimation error is a minimum, and is output to the summing unit 37. The search unit 33 outputs to the half-pixel control unit 34 the present within have similar errors as a re motion estim corresponding adjacent frame integer-pixel ation errors (P.i,0 / P-l', Pl') io the reference image segment within an determined by the motion vector MV1 in units and a number of image segments surrounding thd reference image segment and shifted by one pixel. Advar.tageously, the apparatus of Figure 3A simultaneously and individually determines the horizontal and vertical units. How controller 35 use the same a method for motion vector components of motion vector in half-pixel ever, since the horizontal half-pixel and the vertical half-pixel controller 36 rlnethodology, description will be limited to determining the horizontal component of a in half-pixel units.In the case of the uni-directional motion prediction, a single set of motion estimation errors (P.lzΡ,, Po, P.,', Ρ / ) resulting from comparison of the present frame and adjacent frame is output from the search unit 33 to the half-pixel cor the same argum prediction as set resulting frame and anot and a future f half-pixel cor trol means 34 as described above; however, ent can be made to the bi-directional motion in B-frame, two sets of motion errors, one from comparison of the frame and a previous her set resulting from of the present frame rame, are output from search unit 33 to the trol unit 34.3BFigure adjacent framd In the case o frame are simU is similar to Figure 3A except that two s are simultaneously input to the system. 5 a B-frame, a future frame and a previous Itaneously input to the system.Figure 4 horizontal hal is a detailed block diagram illustrating the f-pixel controllers of Figures 3A and 3B.The devic 41, 42, 43 for Po, P.,) ; a fi input signal I signal Poof a for adding anc input terminal terminal 43 e of Figure 4 includes three input terminals receiving the motion estimation errors (P,, i-st adder Al for adding and outputting the ! of a first input terminal 41 and the input second input terminal 42; a second adder A2 outputting the input signal Poof the second and the input signal P.j of a third input a first comparator CMP1, having a predetermined input terminal coupled to an output terminal of the first adder Al, for comparing respective input signals; a second comparator CMP2, having a predetermined input terminal coupled to an output terminal of the second adder A2, for comparing respective input signals; a second or multiplying the output signal of the by a predetermined coefficient and ltiplied value to the second comparator CMP2; a first multiplier Ml for multiplying the output signal of the second adder A2 by a predetermined coefficient and outputting a multiplied value to the first comparator CMP1; and a NOR gate NOR for performing a unction on respective output signals from CMP1, CMP2. The device further comprises three output tei-minals 44, 45, 46 to distinguish each of the output signals from first comparator CMP1, NOR gate NOR and second comparator CMP2.multiplier M2 f first adder Al outputting a mu, logical NOT OR f the comparatorsThe motion the search unit controller 35 of the first adder estimation errors (PlzPo, P^) output from 33 are input to the horizontal half-pixel the half-pixel control means 34. Then, Al subtracts the signal Poinput through the second terminal 42 input from the input signal P, through the input first terminal 41. The second adder A2 subtracts the signal Poinput through the second input terminal 42 from the signal P., input through the third input terminal 43. The first and second multipliers Ml, M2 multiply by respective predetermined coefficient each of the adder output signals and output the multiplied result to the comparators CMP1, CMP2, connected to the output of the multipliers, respectively. Advantageously, a difference value (a) between Poand P., and (b) between Poand P, are respectively input to the second comparator CMP2 and the first comparator CMP1. Preferably, the first comparator CMP1 compares and outputs the output signals from the first adder Al and the first multiplier Ml. The second comparator CMP2 compares and outputs the output signals from the second adder A2 and the second multiplier M2 .Figures 5(a) and 5(b) are conceptual diagrams showing a method for determining motion vectors in half-pixel units in the horizontal half-pixel controller 35. Figure 5(a) shows that the motion vector MV2 of a half-pixel unit is present at the left with respect to the motion vector MV1 of a pixel vector MV of unit, and Figure 5(b) shows that the motion half-pixel unit is present at the right with the motion vector MVl of a pixel unit.In Figure is larger than located abo is vector MVl in5(a), when Pj is larger than P., (that is, b a) it indicates that the motion vector MV ut 1 / 2 the distance from the first motion jDixel units to the left. In Figure 5(b), when P., is larger than Pj on the basis of Po(that is, a is larger than b) located about 1 pixel units to indicates that the motion vector MV is / 2 way from the first motion vector MVl, in the right.In Figure P.j on the basi5(a), for the case where P, is larger than 3 of Po, only the output 44 of the first comparator CMP1 is high. Thus, the horizontal halfpixel controller 35 outputs -1 / 2, as the horizontal component value of the second motion vector MV2. In Figure 5 (b) , for the case where P., is larger than Pj on the basis of Po, only the output 46 of the second comparatorCMP2 is high. 35 outputs 1 / 2 second motion v Poare equal to comparators CMPThus the horizontal half-pixel controller as the horizontal component value of the ector MV2. Where P, and P., on the basis of each other the output signals of the two1, CMP2 are both low. Therefore, the NOR gate NOR receives low values and its output signal becomes high. Accordingly, the horizontal component of the second motion vector MV2 becomes 0.The summ. MV2 and adds units, obtain the second mot ing unit 37 receives the second motion vector it to the first motion vector MV1, in pixel by the search unit 33, thereby estimating ion vector MV2 in half-pixel units.ed e appreciated .that a method and an apparatus istimation in accordance with preferred bf the present invention compares image bgments within an adjacent frame with the ant in a present frame to detect a motion Elected scale of resolution; generates error ated in the detection of the motion vectors the selected motion vector.It will ] for motion embodiments signals of s selected segm vector at a s values calcul according toBy vertically comparing the generated error values, a vertical component of the motion vector in the finer scale of resolution is detected. By horizontally comparing thfe generated error values, a horizontal component of resolution embodiments o efficiency in the motion vector in the finer scale of is detected. Accordingly, preferred E the present invention lead to a greater motion estimation by reducing the amount of data which must be processed for the detection of the motion vector in a finer scale of resolution.Although the preferred embodiments of the present invention havb been described with the image segment and the search range having a particular form, the image segment and the search range may be of any shape or size such as in the case in which the image segment and the search range is constructed in accordance with the shapes and sizes of the objects. Furthermore, the comparison of a segment from the present frame may be made with an immediately previous frame or immediately future frame as well as previous or future frames separated from the present frame by more than one frame in time. Furthermore, the comparison may be made to any combination of above menti system in whic oned frames such as B-frame used in the MPEG h more than one adjacent frame are compared.Note that cases with othWhile onl been specificc that numerous departing frorr the present invention can be applied to the er selected scale resolutions as well.y certain embodiments of the invention have lly described herein, it will be apparent modifications may be made thereto without the scope of the invention.

Claims

CLAIMS 1. A me frames and estimati of: constructing image frame and segments of a prev hod for comparing an image signal between adjacent ng a motion of the image, the method comprising the steps a first mage segment representing a portion of present a second image segment having a plurality of image ous image frame with respect to the present image frame and a third image segment having a plurality of image segments of a future spect to the present image frame; ίο image frame with re detecting a generated error valu outputting the comparing said plurality of image segments of said second image segment with the image signal within the first image segment, and simultaneously com oaring said plurality of image segments of said third image segment with the im age signal within the first image segment, and generating a number of error values; irst motion vector at one scale of resolution using the es; error value associated with a reference segment pointed by the first motion vector, and the error values corresponding to the reference segment shifted by one unit of selected resolution respectively in selected direction(s); generating a second motion vector in a finer scale of resolution using the output error values; and summing the first and second motion vectors.

2. A method as claimed in claim 1, wherein said generating step further comprises the sub-steps of: detecting a component of said second motion vector in a first direction using the error value of said reference image segment and the error values of said surrounding image segments which are positioned in said first direction with respect to a centre of said reference image segment; and component of said second motion vector in a second error value of said reference image segment and the error ounding image segments which are positioned in said ith respect to the centre of said reference image segment. detecting a direction using the values of said sum second direction wi 53. A me said second directio tlhod as claimed in claim 2 wherein said first direction and n are horizontal direction and vertical direction respectively. ίο4. A method as claimed in claim 2 or claim 3 wherein said component of said second motion vector in said first direction and said component of said second motion vector in said second direction indicate the location of the image segment corresponding to a minimum error value.

5. A me hod as claimed in any preceding claim wherein said previous frame is an immediately previous frame with respect to said present frame. hod as claimed in any preceding claim wherein said 0parated by a plurality of frames in time from said present6. A me previous frame is s frame.

7. A method as claimed in any one of claims 1-4 wherein said future frame is an immediately future frame with respect to said present frame. nod as claimed in any one of claims 1-4 wherein said future by a plurality of frames in time from said present frame.

8. A metl frame is separated thod as claimed in any preceding claim wherein each of e segments of said second image segment has a segment9. A me said plurality of imag thod as claimed in any preceding claim wherein said one a one pixel-unit and said finer scale of resolution is a half size equal to said first image segment.

10. A me scale of resolution i pixel-unit. is11. An apparatus for comparing an image signal between adjacent frames and estimating a motion of the image, the apparatus comprising: means (31) for receiving the image signal of a present frame and constructing a first image segment; means (32) or receiving the image signal of two adjacent frames for io each received image signal corresponding to an adjacent frame, constructing a second image segment from a previous image frame with respect to the present image frame and a third image segment from a future image frame with respect to the present image frame; search means (33) for receiving the image signal output respectively is from the first image segment constructor, and the second image segment constructor and the third image segment constructor, detecting a first motion vector at one selected scale of resolution units, determining a reference segment within said second image segment and within said third image segment in accordance with said first motion vector, and simultaneously for 2o said second and tnird image segments comparing the image signals of surrounding image segments, which are generated by moving the reference image segment in selected direction(s) by one unit of the selected scale respectively, and the reference image segment, with the image signal of the first image, and outputting error values obtained as a result of respective 25 comparisons; control meani (34) for receiving the errors output from the search means and generating a second motion vector at a finer scale of resolution than the selected scale; and summing me^ins (37) receiving the first and second motion vectors, for adding and outputting them.

12. An apparatus as claimed in claim 11, wherein said control means further comprises: a first contro ler (35) for comparing said error value of said reference 5 image segment witi the error values of said surrounding image segments which are positionec in a first direction with respect to a centre of said reference image segment and detecting a vector component in said first direction of said second motion vector; and a second controller (36) for comparing said error value of said reference io image segment with the respective error values of said surrounding image segments which ars positioned in a second direction with respect to the centre of said reference image segment and detecting a vector component in said second direction of said second motion vector. paratus as claimed in claim 12, wherein said first direction direction are horizontal direction and vertical direction13. An ap| is and said second respectively.

14. An apparatus as claimed in claim 12 or claim 13, wherein said controller (35; 36) father comprises: a first subtracter (A21) for subtracting the error value corresponding to 2o said reference image segment from the error values corresponding to said surrounding image segments which are located to one side of said reference image segment; a second subtracter (A2) for subtracting the error value corresponding to said reference image segment from the error values corresponding to said 25 surrounding image segments which are located to the other side of said reference image segment; a first multiplier (M1) for multiplying a respective first output signal of said second subtracter by a predetermined coefficient, and producing a first signal of said first su a second multiplied multiplied value; a second mibltiplier (M2) for multiplying a respective second output jbtracter by said predetermined coefficient, and producing value; 5 a first comparator (CMP1), receiving said second output signal and said first multiplied value, for comparing respective magnitudes of the received first comparator inputs, thereby providing a first binary signal as a comparison result; and a second comparator (CMP2) receiving said first output signal and said io second multiplied value, for comparing the respective magnitude of the received second comparator inputs, thereby providing a second binary signal as a comparison result; and a NOR gate (NOR), receiving said first and second binary signals from said first and second comparators (CMP1, CMP2), respectively, for is performing a logica NOR function, thereby producing an operated value indicative of said second motion vector.

15. An apparatus as claimed in any one of claims 11-14 wherein said previous frame is an frame.

16. An app previous frame is se frame. immediately previous frame with respect to said present ι aratus as claimed in any one of claims 11-14 wherein said oarated by a plurality of frames in time from said present17. An apparatus as claimed in any one of claims 11-14 wherein said future frame is an immediately future frame with respect to said present frame. 2518. An app future frame is sepa aratus as claimed in any one of claims 11-14 wherein said ated by a plurality of frames in time from said present frame.

19. An second image segrr apparatus as claimed in any one of claims 11-18 wherein said ent has a plurality of image segments.

20. An ap 5 of image segments said first image seg >aratus as claimed in claim 19 wherein each of said plurality >f said second image segment has a segment size equal to nent.

21. An apparatus as claimed in any one of claims 11 -20 wherein said one scale of resolution is a one pixel-unit and said finer scale of resolution is a half pixel-unit.