WO2010066801A1

WO2010066801A1 - Method for estimating movement at edges of images

Info

Publication number: WO2010066801A1
Application number: PCT/EP2009/066751
Authority: WO
Inventors: Peter Rieder; Günter Scheffler; Christian Tuschen; Marko Hahn; Markus Schu
Original assignee: Trident Microsystems (Far East) Ltd.
Priority date: 2008-12-10
Filing date: 2009-12-09
Publication date: 2010-06-17
Also published as: DE102008054503A1

Abstract

What is described is a method for testing a motion vector of a first image block (B1 (i) ) of a first image (F (i) ) of an image sequence, said vector located in a first edge area (R1) of the first image (F (i) ), wherein the first edge area (R1) abuts a first edge of the image and wherein the motion vector (vi) comprises a first vector component (v1_x) running perpendicular to the first edge and a second vector component (v1_y), wherein the method comprises: determining a second image block (B2 (i) ) in the first image (F (i) ), said second image block being shifted relative to the first image block (B1 (i)) in the direction running parallel to the first edge (R1) by the first vector component (v1x) in a direction away from the first image edge when the first vector component has a pre-defined direction; determining a third image block (B3 (i) ) in an image that is chronologically in advance of (F(i-1)) or after (F(i+1)) the first image (F (i) ), said third image block being shifted relative to the second image block by the first motion vector (vi); calculating a distance dimension between the contents of the second and those of the third image block B2 (i), B3 (i).

Description

description

METHOD FOR EVOLVING MOTION ESTIMATION OF IMAGING DIMENSIONS

The present invention relates to a method for estimating the motion at image edges.

In motion picture estimation, a motion estimation comprises the determination of motion information for individual picture blocks of temporally successive pictures of a picture sequence. The motion information for an image block of a first image indicates in which direction the image content of the image block shifts between the first image and a temporally preceding or temporally subsequent second image. This motion information is usually reproduced in the form of so-called motion vectors. The motion information obtained by the motion estimation can be used for various purposes, such as e.g. inter-frame interpolation, inter-line interpolation, compression of image data, etc.

There are motion estimation methods, such as recursive motion estimation methods, in which a number of test vectors or candidate vectors are assigned to the individual picture blocks whose motion information is to be determined. For each of the test vectors associated with a first image block, a block matching procedure is performed to determine a distance measure for each of these test vectors. The distance measure for a test vector is a measure of the match between the image content of the first image block in the first image and a second image block in the second image. The position of the second image block is shifted by the test vector with respect to the position of the first image block. As the motion vector for the first image block, the test vector is selected in such a method, for which the "best" distance measure is determined. was telt. As test vectors for the first image block, in the case of recursive estimation methods, for example, the motion vectors of those image blocks are selected which are arranged spatially and / or temporally adjacent to the first image block.

If the first image block is near an image edge, depending on the orientation and the length of the test vector, scenarios may occur in which the second image block lies outside the image or in which the test vector points beyond the image edge. The determination of a distance measure to the test vector is not possible in this case.

WO 2007/119198 describes a method for estimating motion at image edges. In this method, it is determined whether a test vector to be tested for a first image block of a first image points to a second image block that is outside the second image. If so, the block matching method is applied to a shifted first image block and a shifted second image block which are offset from each other by the test vector and which are shifted from the position of the first image block such that the second image block is within the second image. In this method, a shift takes place in particular so far that the second image block lies just within the second image.

In the known method, there is the danger that image blocks that lie in the edge region of the second image are disproportionately taken into account in the motion estimation.

The object of the present invention is to provide an improved method for a motion estimation or for testing motion vectors for picture blocks in the edge region of an image. This object is achieved by a method according to claim 1. Embodiments and developments are the subject of dependent claims.

The invention relates to a method for testing a motion vector to a first image block of a first image of a sequence of images, which is arranged in a first edge region of the image, wherein the first edge region adjoins a first edge of the image and wherein the motion vector is a first perpendicular to comprising the first edge vector component and a second vector component. The method includes: evaluating the first vector component of the motion vector and determining a second image block in the first image shifted from the first image block by the first vector component in a direction away from the first image edge when the first vector component has a first direction ; Determining a third image block in a temporally preceding or temporally subsequent image of the first image, compared to the second image

Image block is shifted by the first motion vector; Calculating a distance measure between the contents of the second and third image blocks.

The determination of the second image block which lies in the first image and which is shifted with respect to the first image block, and performs an image comparison between the second image block and the third image block which lies in the second image and which is opposite the second image block around the test vector - Gate arranged offset, takes place only depending on the direction of the first vector component, but not dependent on the length of this first vector component. The second image may be temporally before or after the first image. If the second image lies temporally before the first image, the first image direction runs away from the first image edge; and if the second image is temporally after the first image, then the first image direction is toward the first image edge.

The image may have, in addition to the first edge region, a second edge region which adjoins a second edge running perpendicular to the first edge. For

Image blocks that lie in this second edge region, the second vector component of the test vector is evaluated.

Embodiments of the present invention will be explained in more detail with reference to figures. The figures serve to explain the basic principle of the invention, so that only the aspects necessary for understanding this basic principle are shown in the figures. The figures are necessarily to scale. In the figures, unless otherwise indicated, like reference numerals refer to like aspects having the same meaning.

Figure 1 illustrates basic aspects of various motion estimation methods.

Figures 2A, 2B illustrate an example of a

Motion estimation method for an image block arranged in a first edge area of an image.

FIG. 3 illustrates individual method steps of a method according to the invention on the basis of a sequence diagram. FIG. 4 illustrates details of the method according to FIGS. 2A, 2B with reference to a table.

FIGS. 5A and 5B illustrate a first example of a motion estimation method for one

Image block arranged in a first and a second edge area of an image.

FIG. 6 illustrates details of the method according to FIGS. 5A, 5B with reference to a table.

Figures 7A and 7B illustrate a second example of a motion estimation method for an image block located in a first edge area of an image.

Figures 8A and 8B illustrate a second example of a motion estimation method for an image block arranged in a first and a second edge area of an image.

FIG. 9 illustrates details of the method according to FIGS. 8A, 8B with reference to a table.

FIG. 1 schematically shows temporally successive images F (il), F (i), F (i + 1) of an image sequence. For further explanation of this image sequence, the image F (i) at the time position i is considered in more detail. This image F (i) is divided into a number of image blocks which are arranged in a grid-like or matrix-like manner. Each of these image blocks comprises a number of pixels arranged in a matrix, at least one of each of these pixels Pixel value or pixel value is assigned. Such a pixel value is, for example, a brightness value (luminance value) or a color value (chrominance value). An image block includes, for example, 8x8 pixels.

The positions of the individual image blocks within the image F (i) are determined by first and second coordinates, which are also referred to below as horizontal and vertical coordinates or x and y coordinates. For purposes of explanation, assume that the x coordinates in the individual images increase from left to right and that the y coordinates increase from top to bottom, so the origin of this coordinate system is in the upper left corner of the individual images.

Bl '(i) designates in FIG. 1 by way of example a first image block in the first image F (i) to which a motion vector vi is to be tested. The testing of such a motion vector vi can in principle be carried out by comparing the image content of the first image block Bl ¹ (i) in the first image F (i) with the image content of a second image block which is contained in the temporally preceding image F (il). is arranged and which is arranged offset from the first image block Bl '(i) around the test vector vi. Such a method is also called backward estimation.

The position of the second image block B2 '(i-1) is at the starting point or foot point of the motion vector vi projected on both images, and the first image block Bl' (i) is located at the end point or the top of the motion picture. the projected motion vector vi. As shown in FIG. 1, there are scenarios in which such a comparison of the image contents of two image blocks is not possible, namely if the first image block Bl ¹ (i) is arranged in the region of an edge of the first image F (i) and if the orientation and the length of the test vector vi are such that it extends beyond the image edge starting from the first image block or that the second image block is arranged outside the second image F (il).

In order to test a test vector vi to an image block of the image F (i), the image content of this image block can also be compared with an image block of the temporally subsequent image F (i + 1). B3 * (i) designates another image block in the first image F (i) in FIG. 1, and v2 designates a test vector to be tested for this image block B3 '(i). In such a method, also referred to as the forward estimation method, the image block B3 '(i) in the image F (i) and the image block B4' (i + 1) in the image F (i + 1) are around the test vector v2 offset from each other, the image block BB ¹ Ci) in the image F (i) at the base or starting point of the test vector v2 and the image block B4 '(i + D in the image F (i + 1) at the end point or the top of the As shown in Figure 1, even in such forward estimation there are scenarios in which the determination of a distance measure for the test vector v2 is not possible, namely when the image block B3 * (i) of the image F (i) in is an edge region of the image, and if an orientation and a length of the test vector v2 are such that the test vector points beyond the edge in the temporally following image F (i + 1) or to a block position outside the temporally following image F (i +1) shows.

Referring now to Figs. 2 to 6, modified motion estimation methods will be explained below which enable a motion estimation also for image blocks in peripheral areas of images. Figures 2A and 2B illustrate a first example of such a method. In FIG. 1, F (i) denotes a first image of an image sequence which has a number of matrix-like or grid-like ordered image blocks. F (il) denotes a second image temporally preceding the first image F (i). The first image F (i) has a first edge region RB1, which is arranged adjacent to a first edge Rl of the first image F (i). In the example illustrated in FIGS. 2A, 2B, this first edge R1 is the left vertical edge of the image, that is to say the edge which bounds the image F (i) to the left in the vertical direction. The choice of this left vertical edge is to be understood in the context of the following explanation only as an example. The following explanations apply correspondingly to edge regions along each of the image edges.

The first edge region RB1 runs parallel to the first edge Rl. Below a width of the edge region RB1, the dimension of the edge region RB1 in a direction perpendicular to the first edge R1 - in the illustrated example, ie in the horizontal direction of the image F (i) - is hereinafter referred to. The width of the edge area RBl is fixed. This width of the edge region RB1 is selected, for example, as a function of the maximum expected length of the test vectors to be tested. This width is for example between 5 and 15 image blocks or between 5-8 and 15-8 pixels, if the individual image blocks have a size of 8x8 pixels.

Bl (i) denotes in FIGS. 2A, 2B a first image block lying within the first edge region RB1, vi denotes a test vector vi to be tested for this first image block Bl (i). This test vector has two

Vector components: A horizontal vector component vl _x , also referred to below as the x-vector component; and - S -

a vertical vector component vl _y, hereinafter referred to as a y vector component. In the example illustrated in FIGS. 2A, 2B, both vector components of the first test vector vi are not equal to zero. Of course, one or both of these vector components may also be zero. In the latter case, the first test vector vi is a zero vector.

The testing of the first test vector vi comprises the determination of a distance measure for this first test vector vi, wherein the determination of a distance measure comprises a comparison of the image contents of image blocks in the first image F (i) and the second image F (i-l). Individual process steps for determining these image blocks to be compared are illustrated in a flowchart in FIG.

In this method, it is provided in a first method step 101 to evaluate the direction of the vector component of the test vector vi, which runs perpendicular to the first edge Rl. This vector component, which runs perpendicular to the first edge Rl, is in the illustrated example the horizontal vector component vl _{x of} the first test vector vi. The evaluation of this vector component v 1 _x comprises evaluating a direction of this vector component as to whether this vector component has a predetermined direction with respect to the first edge. In the illustrated example, such a given direction exists when this vector component points away from the first edge Rl into the image.

Based on the nomenclature used hitherto, the horizontal vector component vl _{x of} the first test vector in the illustrated example then points from the left vertical one Edge Rl away, if this first vector component vl _{x has} a positive sign. If a vector component of a test vector that is perpendicular to a given edge has the predetermined direction, this will be referred to hereinafter as the "edge region violation" with respect to the given edge.

Whether or not there is an edge region violation for an image block Bl (i) arranged in an edge region RB1 and a test vector vi with respect to an edge Rl is dependent on a number of factors, namely the direction of the vector component vlx of the test vector vi and x, which runs perpendicular to the edge Rl the estimation direction, ie the temporal position of the second image F (i-l) with respect to the first image F (i). A backward direction estimate is when the second image F (il) is ahead of the first image and an estimate in a second direction opposite to the backward direction exists if the second image F (i + 1) is later than the second first picture lies. Various scenarios for this will be explained below.

If this vector component vl _x running perpendicular to the first edge Rl has the predetermined direction, ie if there is an edge region violation with respect to the first edge R1, then in a next method step 102 a second pixel B2 (i) within the first image F (i) is determined which is shifted from the first pixel Bl (i) by the first vector component vl _x such that the second image block B2 (i) is located farther apart from the first edge Rl than the first image block Bl (i). The first

Pixel Bl (i) lies at the foot point or at the starting point of the first vector component vl _x , and the second image block B2 (i) lies at the end point or the peak of this vector component vl _x . For purposes of explanation, assume that (xl, yl) are the coordinates of the first image block Bl (i). For the coordinates (x2, y2) of the second image block B2 (i), then:

(x2, y2) = (xl + vl _x, yl) (1).

In a next method step 103, it is provided to determine a third image block B3 (i-1) in the second image F (il) whose position relative to the position of the second image block B2 (i) is offset by the test vector vi such that the the second image block B2 (i) lies at the top or end point and the third image block B3 (i-1) lies at the starting point of the test vector vi, or that the third image block is closer to the edge R1 than the vector component vl _x second image block B2 (i). For coordinates (x3, y3) of the third image block B3 (i-1), the following applies with respect to the coordinates of the second image block:

(x3, y3) = (x2-vl _x, y2-vl _y) = (xl, yl-vl _y) (2).

The horizontal coordinate x3 of the third image block B3 (i-1) corresponds to the horizontal coordinate of the first image block Bl (i).

In this context, it should be noted that the test vector vi in the example shown has a pixel-precise resolution, ie a resolution which is higher than the resolution of the block raster. Accordingly, neither the second image block B2 (i) nor the third image block B3 (i-1) must lie on the block grid, which is shown for better understanding in Figs. 2A, 2B for both images F (i), F (il). The test vector vi can not be shown in detail Even a sub-pixel-accurate resolution, that is, a resolution that is higher than the image pixel (pixel) grid of the image.

In a next method step 104, referring to FIG. 3, a distance for the test vector vi is determined by comparing the contents of the second and third image blocks B2 (i), B3 (i-1). Thus, image blocks are compared with each other for whose coordinates, referring to the equations (1) and (2): (xl + vl _x , yl), (xl, yl-vly). In contrast, if there were no edge region violation, the contents of the first image block Bl (i) with the coordinates (xl, yl) and another image block, which is shifted from the first image block by the test vector vi, would be directly in the second image B2 (i-1) and having the coordinates (xl-vl _x , yl-vl _y ), compared with each other.

The determination of the distance measure can be done in any basically known manner. In one example, it is provided to form the differences of the pixel values of pixels located at the same positions within the two blocks B2 (i), B3 (i-1), to form the absolute values of these obtained differences, and to accumulate the thus obtained absolute values , The value thus obtained is also called the sum of the absolute values of the pixel value differences. Instead of the absolute values, any other metrics may be applied to the differences of the pixel values, such as even powers.

In such a determination of the distance measure, the distance measure obtained is the smaller, the more the contents of the two image blocks match. As pixel values for the comparison of the image contents of the two image blocks, any of the pixel values assigned to the individual pixels can be used, For example, brightness values (luminance values) or color values (chrominance values).

Is the orientation of the perpendicular to the edge Rl vector component of the test vector such that it is opposite to the predetermined orientation, so that to the edge Rl perpendicular vector component vl _x points in the direction of the edge, as a conventional reverse estimate can be made, as illustrated for a wide ren in the first edge region RBl arranged image block B4 (i). In FIGS. 2A, 2B, v2 denotes a to this further

Image block B4 (i) test vector v2 to be tested. A vector component v2 _{x of} this test vector v2 running perpendicular to the edge Rl shows in the example in the direction of this edge R1. A distance measure for this test vector v2 can thus be determined directly using the image content of the further image block B4 (i) and an image block B5 (il) in the second image F (il). The image block B5 (i-1) in the second image F (il) is offset by the test vector v2 from the position of the image block B4 (i). (x4, y4) let be the coordinates of the image block

B4 (i) in the first image F (i). For the coordinates (x5, y5) of the image block B5 (i-1) in the second image F (i-1), then:

(X5, y5) = (x4 _x v2 _/ v2 _y4-y) (3).

v2 _x , v2 _y denote the vertical and horizontal vector component of the test vector v2.

The above-explained method for testing a test vector for an image block, such as the first image block Bl (i) according to FIGS. 2A, 2B, can be applied to a plurality of different test vectors, wherein the For example, the distance measures obtained for the individual test vectors are compared with one another, and one of these test vectors is selected as a motion vector for the image block taking into account a distance measure. The group of test vectors to be tested for an image block may be selected in a conventional manner. Test vectors for an image block, such as the first image block Bl (i), are, for example, motion vectors previously arranged temporally and / or spatially adjacent to the first image block Bl (i)

Image blocks were determined. Spatially adjacent image blocks are image blocks that are arranged within the first image F (i) spatially adjacent to the first image block Bl (i). Time-adjacent image blocks are image blocks located at the same position as the first image block Bl (i), but arranged in temporally adjacent images, such as the second image F (I-I). Temporally and spatially adjacent image blocks are image blocks located in images other than the first image F (i) at positions spatially adjacent to the position of the first image block Bl (i).

As already mentioned, the explained method can be applied to image blocks in arbitrary edge regions of the image F (i), that is to say edge regions along all four edges of the image. The method is also applicable to edge regions within the image, for example edge regions along edges between different image regions of the image in which independent image information from different image sources is displayed. The representation of independent image information in different image areas is known in principle. Examples include: a picture-in-picture overlay, where one picture in a window in another Picture is presented; a so-called on-screen display (OSD), ie the presentation of menu information in a menu window, which is displayed in an image; or a "split-screen" view, in which the screen is divided into several windows, each of which displays independent image information.

Regardless of which edge region the first image block Bl (i) lies in, the orientation of the vector component of the test vector perpendicular to the edge is always evaluated, and a second image block B2 (i) arranged in the first image F (i) is then determined when the orientation of this perpendicular vector component is a given orientation, ie based on the example discussed so far, when this vector component points away from the edge. In the table reproduced in FIG. 4, the vector components to be evaluated and the sign of this vector component are reproduced for the outer four edges of the image in the case that this vector component points away from the respective edge. The sign in this table for the individual vector components refers to the previously used coordinate system for the block grid, in which the origin of the coordinate system lies in the upper left corner of the image.

The border areas running along the outer edges of the image overlap in the corners of the image. For pixels which are arranged simultaneously in two different overlapping edge regions of the image, the orientations of the vector components running perpendicular to the edges are evaluated independently of one another and the result obtained is taken into account for the determination of the position of the second image block B2 (i) , There are four different cases, which are explained below. In this explanation, "first

Vector component "a perpendicular to the first edge running Vector component of the test vector, and "second vector component" denotes a vector component of the test vector that is perpendicular to the second edge:

a. The second image block matches the first image block if none of the first and second vector components point away from the respective edge;

b. The position of the second image block is offset by the first vector component from the position of the first image block when only the first vector component is pointing away from the first edge;

c. The position of the second image block is offset from the position of the first image block by the second vector component if only the second vector component points away from the second edge;

d. The position of the second image block is offset from the first image block by the first and second vector components when both the first vector component is pointing away from the first edge and the second vector component is pointing away from the second edge.

FIG. 5 illustrates a test method for a first image block Bl (i) which is arranged simultaneously in two edge regions of the image: a first edge region RB1, which is arranged adjacent to a first edge R1; and a second edge region RB2 disposed adjacent to a second edge R2 of the image F (i). The first edge R1 is the left vertical edge of the image F (i) in the illustrated example, and the second edge R2 is the bottom horizontal edge of the image F (i) in the illustrated example. One through these two edge regions RBl, RB2 defined corner region of the image F (i) is a left lower corner region of the image F (i). The selection of this corner area for the following explanation is merely an example. The following explanations apply correspondingly to any corner region of the image F (i) defined by two edge regions.

vi denotes a test vector in Figures 5A, 5B, vl _x denotes the horizontal vector component and lr _y denotes the vertical vector component of this test vector vi. These two vector components vl _x , vl _y are not equal to zero in the illustrated example. Of course, however, there is also the possibility that one of these vector components or that both of these vector components are equal to zero.

The direction perpendicular to the first edge Rl first vector component of the in Figures 5A, test vector 5B vi, is the horizontal vector component vl _x, and the direction perpendicular to the second edge R2 second vector component of the test vector vi is in the example shown, the vertical vector component vl _y . to

Determining the position of the second pixel B2 (i), the orientation of the first and the second vector component are evaluated, wherein there are four different cases according to the previously made embodiments here. These case distinctions are shown in FIG. 6 in the form of a table.

Shown in FIG. 6 is the difference between the positions of the second image block B2 (i) and the first image block Bl (i), ie the spatial offset between these two image blocks. This spatial offset is dependent on the orientation of the first and second vector components, or the sign of these two Vector components. If the first vector component Vl _{x has} a positive sign, and the second vector component vl _{y has} a negative sign, then in the example both vector components point away from the respective edges. The second pixel B2 (i) is in this case both the first and the second vector component vl _X; vl _y shifted from the position of the first image block Bl (i). Having the first vector component vl _x has a positive sign and the second vector component vl _y e-benfalls a positive sign, the second image block B2 is (i) QUIRES ONLY lent to the first vector component vl _x relative to the position of the first image block Bl (i) postponed. If the first vector component vl _{x has} a negative sign and the second vector component vl _{y has} a positive sign, the second image block B2 (i) is shifted from the position of the first image block Bl (i) by the second vector component vl _y . Having the first vector component vl _x in the example, a negative sign and the second vector component vl _y has a positive sign, then both vector components point in the direction of the respective edge. In this case, no second image block B2 (i) within the first image F (i) is determined, or the second image block in this case corresponds to the first image block Bl (i), ie the displacement is (0, 0).

The third image block B3 (i-1) in the second image F (i-1) is shifted by the value of the test vector vi from the position of the second image block B2 (i) in accordance with the explanations given above. In the example shown in Figs. 5A, 5B, in which the second image block B2 (i) is shifted from both the first vector component and the second vector component from the position of the first image block Bl (i), the position of the third one

Image block B3 (i-1) of the position of the first image block Bl (i). In both methods explained above, for a first image block Bl (i) within a border region, a second image block B2 (i) within the same image is determined, the position of this second image block B2 (i) being dependent on the orientation of a plane perpendicular to the border Vector component of the test vector and is dependent on the magnitude of this vector component.

The method explained above may be referred to as a modified backward estimation method, since in this method for determining motion information about an image block-in the example the first image block Bl (i) -of the first image F (i), image information from a temporally preceding image block second image F (il) can be used.

The above-explained method steps can also be correspondingly applied to a forward estimation method, in order thereby to obtain a modified forward estimation method, as will be explained below with reference to FIGS. 7 and 8. Image information of image blocks of the first

Image F (i) are compared in this method with image blocks in a second image F (i + 1), which follows in time the first image F (i). Such a method differs from the method explained above in that when testing a motion vector to a first image block in the first image F (i), a second image block whose position differs from the position of the first image block is then determined when the second image block first image block is arranged in an edge region adjacent to an edge and if the vector component of the test vector running perpendicular to this edge points in the direction of this edge. An example of such a method is explained below with reference to FIGS. 7A, 7B. Bl (i) in FIGS. 7A, 7B denotes a first image block of the image F (i), which is arranged in a first edge region RB1. This first edge region RB1 is arranged adjacent to a first edge R1, which in the illustrated example is a right vertical edge of the image F (i). vi denotes the test vector to be tested for this image block Bl (i), and vl _x denotes the vector component of this test vector which is perpendicular to the first boundary Rl. This vertically ver running vector component vl _x runs in the illustrated example in the direction of this first edge Rl. In this case, a second image block B2 (i) whose position is shifted relative to the position of the first image block Bl (i) by the vector component vl _x perpendicular to the edge Rl is determined in the first image F (i). The second image block B2 (i) lies at the starting point and the first image block Bl (i) lies at the end point of this vector component. Starting from the position of the second image block B2 (i), a third image block B3 {i) is subsequently determined in the second image F (i + 1), the position of the third image block B3 (i) being opposite the position of the test vector vi second image block B2 (i) is shifted, wherein the second image block B2 (i) at the start point and the third image block B3 (i) at the end point of the test vector vi. For the coordinates (x2, y2) of the second image block B2 (i) with respect to the coordinates (xl, yl) of the first image block Bl (i) and for the coordinates (x3, y3) of the third image block B3 (i) relative to the Coordinates of the second image block B2 (i) apply in the example shown:

(x2, y2) = (Xl-Vl _x, yl) (4)

(x3, y3) = (x2 + _x vl, v2 y2 + _y) (5). Using the second and third image blocks B2 (i), B3 (i + 1), a distance measure for the test vector vi is then determined. To determine the distance measure any conventional and previously discussed methods are suitable.

If the vector component of the vector to be tested, which is perpendicular to the edge R 1, points away from the edge R 1, as shown for the further image block B 4 (i) shown in the first edge region RB 1, then the distance measure for the test vector, in the example shown, becomes Vector v2, using the image content of this image block B4 (i) and another image block B5 (i + 1) in the second image F (i + 1). The image block B5 (i) in the second image F (i + 1) is offset by the test vector v2 from the position of the image block B4 (i), the image block B4 (i) in the first image F (i) at the starting point and the image block B5 (i + 1) is at the end point of the vector v2 to be tested.

In the example shown in Figures 7A, 7B, the perpendicular to the edge Rl component vector is the horizontal vector component _x vl. This vector component points to the first edge R1 - the right vertical edge in the illustrated example - if this vector component has a positive sign. The explained method is, of course, applicable in a corresponding manner to any edge regions of the image F (i). The vector components to be evaluated for the individual edge regions and their signs for the case in which this vector component points to the edge are shown in the right-most column of FIG. For first image blocks, which are located simultaneously in two different edge regions, the vector components of the vector to be tested, which are perpendicular to the two edges, are evaluated separately, as shown in FIGS. 5 and 6, and the position of the second image block B2 (i) is determined according to FIG Explanations to Figures 5A, 5B depending on whether one, both or none of these vector components point in the direction of the respective edge. An example of a method for testing a test vector for a first image block Bl (i) simultaneously in two

Is illustrated in FIGS. 8A, 8B, vi denotes a test vector vi to be tested for this image block Bl (i) in FIGS. 8A, 8B. The first image block Bl (i) lies at the same time within a first edge region RB1 and a second edge region RB2. In this case, the first edge area RB1 is adjacent to a first edge R1, in the example the right vertical edge, and the second edge area RB2 is arranged adjacent to a second edge R2, in the example the upper horizontal edge. The direction perpendicular to the first edge Rl vector component is in the example shown, the horizontal vector component _x vl, and the direction perpendicular to the second edge vector component R2 is the vertical vector component vl _y. Both of these vector components point in the example shown in the direction of the respective edge. The second image block B2 (i) is shifted from the position of the first image block Bl (i) by the first and second vector components. With regard to the position of this second image block B2 (i), depending on the orientation of these vector components, there are four case distinctions, which are illustrated in accordance with the embodiments of FIGS. 5 and 6 in FIG.

If, in the method explained so far, there is an edge area violation with respect to a first image block Bl (i) and one of these In the manner explained above, a second image block B2 (i) is determined in the first image F (i), in which the first image block Bl (i) is arranged, in the image block Bl (i) to be tested , which is shifted from the position of the first image block Bl (i). An image comparison is then performed using this second image block B2 (i) from the first image F (i) and a third image block B3 (i-1) from the second image Fl (il), the position of the third image block B3 (i -1) is shifted by the test vector vi from the position of the second image block. However, strictly speaking, the test result obtained by the image comparison no longer refers to the first Bl (i) but to the second image block B2 (i).

However, in order to test a motion vector vi for a first image block B2 (i) of the first image, not only a temporally preceding F (il) or a temporally succeeding F (itl) image is available, but both of these images are available another way to respond to a border violation, which is explained below:

If there is an edge region violation for a test vector to an estimation direction, then it is intended to carry out an estimation in the opposite estimation direction. This will be explained below with reference to the examples in FIGS. 2 and 7. Referring to the previous explanation, in the example of FIGS. 2A, 2B, a backward estimation is performed. For the test vector vi and the first image block Bl (i), there is an edge region violation. In order to be able to perform a motion estimation despite this edge area violation, the content of the image block Bl (i) becomes the content of an image block B5 (i + 1) in the temporally successive image F (i + 1) whose coordinates (x5, y5) are valid for:

(x5, y5) = (xl + vlx, yl + vly) (6).

Thus, a forward estimate, i. an estimate carried out in an opposite direction.

Referring to the previous explanation, in the example of FIGS. 7A, 7B, a forward estimation is performed. For the test vector vi and the first image block Bl (i), there is an edge region violation. In order to be able to perform a motion estimation despite this edge area violation, the content of the image block Bl (i) is compared with the content of an image block B5 (i-1) in the temporally preceding image F (il), whose coordinates (x5, y5) apply :

(x5, y5) = (xl-vlx, yl-vly) (6).

Thus, a backward estimate, i. an estimate carried out in an opposite direction. Generally speaking, for motion estimation in this method, a second image block B5 (i) is selected which is shifted from the first image block Bl (i) by the first motion vector vi and which depends on whether the first vector component vlx is a predetermined one Having direction in a first image F (i) temporally preceding image F (i-1) or one of the first image F (i) temporally subsequent image F (i + 1) is arranged.

If the block lies at the same time in two edge regions, that is to say in one of the corners of the image, the case may arise that a margin violation for one of the components of the vector and a margin violation for one of the components of the vector in a forward estimation. In this case, for example, wise recourse to the explained with reference to Figures 5 or 8 method. In addition, for blocks located in a corner, it is possible to resort to the method according to FIGS. 5 or 8, in principle, if there is a margin violation, even if an estimation in the opposite direction is possible, ie without margin violation.

Claims

claims

A method of testing a motion vector to a first image block (Bl (i)) of a first image (F (i)) of an image sequence located in a first edge region (Rl) of the first image (F (i)) the first edge region (R) adjoins a first edge of the picture and the motion vector (vi) a first perpendicular to the first edge running vector component (vl _x) and a second vector component (vl _y), said method comprising:

Determining a second image block (B2 (i)) in the first one

Image (F (i)) shifted from the first image block (Bl (i)) in the direction perpendicular to the first edge (Rl) by the first vector component (vlx) in a direction away from the first image edge the first vector component has a predetermined direction;

Determining a third image block (B3 (i)) in a first (F (i)) temporally preceding (F (il)) or temporally subsequent (F (i + 1)) image, which is compared to the second image block by the first Motion vector (vi) is shifted;

Calculating a distance measure between the contents of the second and third image blocks B2 (i), B3 (i).

2. A method of testing a motion vector for a first image block (Bl (i)) of a first image (F (i)) of an image sequence arranged in a first edge region (R1) of the first image (F (i)) the first edge region (R) adjoins a first edge of the picture and the motion vector (vi) a first perpendicular to the first edge running vector component (vl _x) and a second vector component (vl _y ), the method comprising:

Determining a second image block which is displaced relative to the first image block Bl (i) by the first motion vector (vi) and which depends on whether the first vector component (vlx) has a predetermined direction in a first image (F (i) ) temporally preceding image (F (il)) or a first image (Bl (i)) temporally succeeding image (F (i + 1)) is arranged.

The method of claim 1 or 2, wherein the predetermined direction is such that the first vector component perpendicular to the first edge faces away from the first edge (R 1) into the image and the second image represents the temporally preceding image is.

4. The method of claim 1, wherein the predetermined direction is such that the first vector component perpendicular to the first edge points from the image to the first edge and wherein the second image is the temporally subsequent image ,

5. The method according to any one of the preceding claims, wherein the first image block Bl (i) is arranged simultaneously in a second edge region (RB2), which adjoins a second edge (R2) and wherein the second vector component (vl _y ) perpendicular extending to this second edge (R2) and further comprising:

Determining the second image block (B2 (i)) in the first image (F (i)) such that this image block is in the direction perpendicular to the second edge (R2) about the second vector component (vly) is offset from the position of the first image block (Bl (i)) when the second vector component has a predetermined direction.

6. The method of claim 4, wherein the predetermined Rieh tion is such that the second perpendicular to the second edge (R2) extending vector component from the second edge (R2) away into the image and in which the second image in time previous picture is.

The method of claim 5, wherein the predetermined direction is such that the second vector component perpendicular to the second edge (R2) points away from the second edge (R2) into the image, and wherein the second image is time-wise subsequent picture is.

8. The method according to any one of the preceding claims, wherein the images (F (i), F (il)), F (i + 1)) of the image sequence are displayed in a picture-in-picture representation in images of another image sequence ,

9. Method according to one of the preceding claims, in which the images (F (i), F (i-1)), F (i + 1)) of the image sequence are displayed in a "split-screen" representation in addition to images of at least one further image sequence.