Image analysis – Image compression or coding – Interframe coding
Reexamination Certificate
1998-04-14
2001-05-29
Mancuso, Joseph (Department: 2621)
Image analysis
Image compression or coding
Interframe coding
Reexamination Certificate
active
06240211
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a method for motion estimated and compensated Field Rate Up-conversion (FRU) for video applications, and to a device for actuating such method More particularly, the invention relates to a method based on a so-called “block-matching technique”.
2. Discussion of the Related Art
The market introduction of high-end TV sets, based on 100 Hz Cathodic Ray Tubes (CRTs), required the development of reliable Field Rate Up-conversion (FRU) techniques to remove artifacts such as large area flicker and line flicker.
Methods for FRU are known (hereafter referred to as “standard FRU methods”) that add interpolated missing image fields to a video source. The standard methods interpolate the missing fields to be displayed on the CRT without performing an estimation and compensation of motion of moving objects in successive image fields of the video source.
The standard FRU methods are satisfactory for improving the quality of an image and reducing artifacts such as large-area flicker or line flicker in the up-converted output. However, when FRU is performed by means of the standard methods, new artifacts can appear in the displayed image. In particular, if the image contains moving objects, “motion judder” is introduced. In fact, the standard interpolation algorithms are not able to detect an objects' motion, and this can lead to interpolated fields in which the moving objects are displayed in wrong positions.
The problem is better understood by referring to
FIG. 1
, in which the motion trajectory of a moving object (white squares) in the original video source image fields is supposed to be a straight line. If the missing fields are interpolated by means of a standard FRU method (i.e. without motion estimation and compensation), the position of the moving object in the interpolated fields (dark gray squares) is not as expected by a viewer (dotted squares). Such artifacts are visible to the viewer, inducing a blurring effect for fast moving objects and considerably reducing the quality of the displayed images.
In order to avoid such a blurring effect and to reduce artifacts in general, other FRU methods have been proposed that are capable of performing a motion estimation and compensation of moving objects in the image fields. Essentially, motion estimation and compensation provides for detecting the moving parts of the video source image fields and interpolating the missing fields for up-conversion according to the estimated motion. The movement of objects in consecutive image fields can be represented by so-called “motion vectors”.
FIG. 2
shows an example of a known FRU method with motion estimation and compensation, where an image containing moving objects is considered. Between two consecutive source image fields, a moving object may have changed its position; for example, object MO in a previous field (Field T) is in position A and in a current field (Field T+1) is in position B. A motion exists from the previous field to the current field, and this motion can be represented by a vector AB, called a motion vector.
The motion vector AB represents the motion of object MO from position A in the previous field to position B in the current field: starting from position A in the previous field and applying the motion vector AB to object MO, the object MO is translated into position B in the current field. The position I of the object MO in the intermediate missing field to be interpolated (Field T+½) must be calculated by the interpolation of the previous field and the current field, taking into account the respective positions A and B of the moving object MO. For example, if object MO changes position between the previous field and the current field, position I in the missing field is obtained by the translation of A with a motion vector |AB|/2. In this way, it is possible to avoid the blurring effect because the missing field is interpolated with the moving object in an appropriate position.
Theoretically, a motion vector could be calculated for each pixel of a field. However, this would require a large number of calculations and substantial memory. In practice, it is assumed that the dimensions of the objects in the source image are always larger than that of a pixel. The image field is therefore divided into image blocks IB (
FIG. 3
) and a motion vector for each block is calculated. The dimension of the image blocks in terms of pixels is generally chosen on an experimental basis. The position of an image block in a field is identified by the coordinates of the first pixel (upper-left) of the block in the field. Typically, it is also assumed that the movement of each image block is rigid and translational only.
Known methods that include motion estimation generally provide for detecting corresponding image blocks in two consecutive source image fields and interpolating the missing fields according to the relative positions of the blocks. As shown in
FIG. 4
, a matrix MV (Motion Vectors) is associated with a pattern of image blocks in a missing field. This matrix MV includes a motion vector for each image block of the pattern. Each block K(x,y) of the missing field has a position x, y, where x and y are the coordinates of the upper-left pixel of the block. The position of the block in the missing field is between the position of a corresponding block B
1
in the previous source image field and a corresponding block B
2
in the current source image field.
Blocks B
1
and B
2
are related by the motion vector V in matrix MV. For a generic block K(x,y) in the missing field, the corresponding vector in matrix MV is a vector V(dx,dy). From the position of the motion vector in the matrix MV and the value of the motion vector, the positions of the blocks B
1
and B
2
in the previous and current fields, respectively, are given by:
B
1(
x−dx;y−dy
) and B2(
x+dx;y+dy
) (1).
Thus, once the matrix MV of motion vectors has been built, the motion of each block between consecutive source image fields is determined, and the missing fields can be interpolated with the blocks in correct positions.
To build up the matrix MV of motion vectors, the image blocks of a missing field are scanned starting from the upper-left down to the bottom-right. According to a so-called “block-matching” technique, for each image block some neighboring blocks are selected, together with their respective motion vectors. For those neighboring blocks that precede the block under examination in the scanning sequence for the missing field, the previously calculated motion vectors for the preceding blocks are used; for those neighboring blocks following the block under consideration in the scanning sequence, for which new motion vectors have not yet been calculated, the motion vectors of a previously calculated and stored matrix MV are used. The motion vector associated with the block under consideration is then calculated on the basis of the motion vectors of the neighboring blocks.
Such a method involves a recursive algorithm, and the performance of the method, as well as the results of the displayed image, depend on the choice of the neighboring blocks and on the way in which the motion vector for a particular block is calculated from the motion vectors of the neighboring blocks.
In view of the state of the art, it is an object of the present invention to provide an improved method for motion estimated and compensated FRU.
SUMMARY OF THE INVENTION
According to the present invention, these and other objects are achieved by a method and apparatus for motion estimated and compensated field rate up-conversion. The invention provides for dividing an image field to be interpolated into a plurality of image blocks, where each image block is made up of a respective set of image elements of the image field.
In one embodiment, for each image block of a subset of image blocks, a group of neighboring image blocks is selected. A motion vector for the image block is estimated that describ
D'Alto Viviana
Mancuso Massimo
Molinari Luca
Poluzzi Rinaldo
Bayat Ali
Galanthay Theodore E.
Mancuso Joseph
Morris James H.
SGS--Thomson Microelectronics S.r.l.
LandOfFree
Method for motion estimated and compensated field rate... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for motion estimated and compensated field rate..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for motion estimated and compensated field rate... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2521034