Subpixel accurate motion vector estimation and motioncompensated interpolation
First Claim
1. A method of estimating subpixel accurate motion vectors between first and second images having a given mutual temporal distance, the subpixel accurate motion vectors being intended for use in an interpolation or prediction of an image at a fractional distance from said first image, said fractional distance being a fraction of said given mutual temporal distance, the method comprising the steps:
 generating a candidate vector having vector components that may have noninteger values;
deriving first and second vectors from said candidate vector; and
evaluating said candidate vector to determine said subpixel accurate motion vectors by comparing a first position in said first image shifted over said first vector to a second position in said second image shifted over said second vector;
characterized in that said deriving step comprises the steps;
multiplying said vector components of said candidate vector by said fraction to obtain fractional vector components;
rounding said fractional vector components to obtain vector components of said first vector, said first vector having only integer vector components; and
subtracting said first vector from said candidate vector to obtain said second vector, whereby said second vector has vector components that, depending on said candidate vector and said fraction, may have noninteger values.
10 Assignments
Litigations
0 Petitions
Accused Products
Abstract
Subpixel accurate motion estimation from digital video requires interpolation of data. In a symmetrical motion estimator that shifts the neighboring images over complementary fractions (−D/2, D/2) of the vector (D), two such interpolating filters are required. In case of blockmatching motion estimation algorithms, the match error of various candidate vectors depends on the quality of the subpixel interpolation filter(s). As these filters differ per fractional value of the vector, unintentional preferences for some fractional vector values over other may result. The candidate vectors (D) are split into an integer part (Rnd(D/2)) and a generally noninteger part (DRnd(D/2)), as this gives a better accuracy and a lower cost than splitting the candidate vectors (D) exactly conform the temporal position at which the vectors need to be valid.
43 Citations
7 Claims

1. A method of estimating subpixel accurate motion vectors between first and second images having a given mutual temporal distance, the subpixel accurate motion vectors being intended for use in an interpolation or prediction of an image at a fractional distance from said first image, said fractional distance being a fraction of said given mutual temporal distance, the method comprising the steps:

generating a candidate vector having vector components that may have noninteger values;
deriving first and second vectors from said candidate vector; and
evaluating said candidate vector to determine said subpixel accurate motion vectors by comparing a first position in said first image shifted over said first vector to a second position in said second image shifted over said second vector;
characterized in that said deriving step comprises the steps;
multiplying said vector components of said candidate vector by said fraction to obtain fractional vector components;
rounding said fractional vector components to obtain vector components of said first vector, said first vector having only integer vector components; and
subtracting said first vector from said candidate vector to obtain said second vector, whereby said second vector has vector components that, depending on said candidate vector and said fraction, may have noninteger values.


2. A method of generating an intermediate image using subpixel accurate motion vectors having vector components that may have noninteger values, from first and second images having a given mutual temporal distance, the intermediate image being at a fractional distance from said first image, said fractional distance being a fraction of said given mutual temporal distance, the method comprising the steps:

deriving first and second vectors from said subpixel accurate motion vectors; and
generating said intermediate image by combining first positions in said first image shifted over said first vectors and second positions in said second image shifted over said second vectors;
characterized in that said deriving step comprises the steps;
multiplying said vector components of said subpixel accurate motion vectors by said fraction to obtain fractional vector components;
rounding said fractional vector components to obtain vector components of said first vectors, said first vectors having only integer vector components; and
subtracting said first vector from said candidate vector to obtain said second vector, whereby said second vectors has vector components that, depending on said candidate vector and said fraction, may have noninteger values.  View Dependent Claims (3)
multiplying said subpixel accurate motion vectors by a factor equal to 1 minus said fraction to obtain said second vectors. 

4. A device for estimating subpixel accurate motion vectors between first and second images having a given mutual temporal distance, the subpixel accurate motion vectors being intended for use in an interpolation or prediction of an image at a fractional distance from said first image, said fractional distance being a fraction of said given mutual temporal distance, the device comprising:

means for generating a candidate vector having vector components that may have noninteger values;
means for deriving first and second vectors from said candidate vector; and
means for evaluating said candidate vector to determine said subpixel accurate motion vectors by comparing a first position in said first image shifted over said first vector to a second position in said second image shifted over said second vector;
characterized in that said deriving means comprises;
means for multiplying said vector components of said candidate vector by said fraction to obtain fractional vector components;
means for rounding said fractional vector components to obtain vector components of said first vector, said first vector having only integer vector components; and
means for subtracting said first vector from said candidate vector to obtain said second vector, whereby said second vector has vector components that, depending on said candidate vector and said fraction, may have noninteger values.


5. A device for generating an intermediate image using subpixel accurate motion vectors having vector components that may have noninteger values, from first and second images having a given mutual temporal distance, the intermediate image being at a fractional distance from said first image, said fractional distance being a fraction of said given mutual temporal distance, the device comprising:

means for deriving first and second vectors from said subpixel accurate motion vectors; and
means for generating said intermediate image by combining first positions in said first image shifted over said first vectors and second positions in said second image shifted over said second vectors;
characterized in that said deriving means comprises;
means for multiplying said vector components of said subpixel accurate motion vectors by said fraction to obtain fractional vector components;
means for rounding said fractional vector components to obtain vector components of said first vectors, said first vectors having only integer vector components; and
means for subtracting said first vector from said candidate vector to obtain said second vector, whereby said second vectors has vector components that, depending on said candidate vector and said fraction, may have noninteger values.  View Dependent Claims (6, 7)
means for multiplying said subpixel accurate motion vectors by a factor equal to 1 minus said fraction to obtain said second vectors. 

7. A television apparatus, comprising:

means for receiving a television signal;
the generating device as claimed in claim 5, coupled to said receiving means, to obtain a display signal; and
means for displaying said display signal.

1 Specification