Motion or depth estimation by prioritizing candidate motion vectors according to more reliable texture information
View Patent ↗In a motion or depth estimation method, a set (CS) of candidate motion vectors, or depth values formed by already obtained motion vectors, or depth values for neighboring image parts which are spatio-temporally adjacent to a given image part of interest, is generated (MV-MEM) for the given image part of interest. Those candidate motion vectors or depth values which correspond to neighboring image parts containing more reliable texture information than other neighboring image parts, are prioritized (TBPD) to obtain a prioritized set (PCS) of candidate motion vectors or depth values. Thereafter, motion or depth data (MV) for the given image part of interest is furnished (ME) in dependence upon the prioritized set (PCS) of candidate motion vectors or depth values. Finally, an image signal processing device (ISPD) processes an image signal (Iin) to obtain an enhanced image signal in dependence upon the motion or depth data (MV).
1. A motion estimation method, comprising the steps:
generating, for a given image part of interest, a set of candidate motion vectors formed by already obtained motion vectors for neighboring image parts which are spatio-temporally adjacent to said given image part of interest;
prioritizing those candidate motion vectors which correspond to neighboring image parts containing more reliable texture information than other neighboring image parts, to obtain a prioritized set of candidate motion vectors; and
furnishing motion data for said given image part of interest in dependence upon said prioritized set of candidate motion vectors.
2. A motion estimating method as claimed in claim 1 , wherein:
said prioritizing step comprises generating a set of N−K candidate motion vectors for N−K blocks from a set of N motion vectors for N blocks including motion vectors for a block of interest and N−1 blocks spatio-temporally adjacent to said block of interest, said N−K candidate motion vectors being formed by excluding K motion vectors for K blocks having lower confidence quantities than confidence quantities for a remaining N−K motion vectors for a remaining N−K other blocks, where N and K are integers and N>K; and
said motion data furnishing step comprises furnishing motion information in dependence on said set of N−K motion vectors.
3. A motion estimating method as claimed in claim 2 , wherein said motion information furnishing step includes:
adding a random value to said set of N−K motion vectors thereby forming a set of N−K+1 motion vectors; and
selecting an optimal candidate from said set of N−K+1 motion vectors.
4. A motion estimating method as claimed in claim 2 , wherein blocks are updated in the order of decreasing confidence quantity.
5. A motion estimating method as claimed in claim 2 , wherein said confidence quantity depends on matching quality and/or informational entropy.
6. A method of extracting depth information from motion, the method comprising the steps:
estimating motion data as claimed in claim 1 ; and
generating depth information from the motion data.
7. A depth estimation method, comprising the steps:
generating, for a given image part of interest, a set of candidate depth values formed by already obtained depth values for neighboring image parts which are spatio-temporally adjacent to said given image part of interest;
prioritizing those candidate depth values which correspond to neighboring image parts containing more reliable texture information than other neighboring image parts, to obtain a prioritized set of candidate depth values; and
furnishing depth data for said given image part of interest in dependence upon said prioritized set of candidate depth values.
8. A motion estimation device, comprising:
means for generating, for a given image part of interest, a set of candidate motion vectors formed by already obtained motion vectors for neighboring image parts which are spatio-temporally adjacent to said given image part of interest;
means for prioritizing those candidate motion vectors which correspond to neighboring image parts containing more reliable texture information than other neighboring image parts, to obtain a prioritized set of candidate motion vectors; and
means for furnishing motion data for said given image part of interest in dependence upon said prioritized set of candidate motion vectors.
9. A depth estimation device, comprising:
means for generating, for a given image part of interest, a set of candidate depth values formed by already obtained depth values for neighboring image parts which are spatio-temporally adjacent to said given image part of interest;
means for prioritizing those candidate depth values which correspond to neighboring image parts containing more reliable texture information than other neighboring image parts, to obtain a prioritized set of candidate depth values; and
means for furnishing depth data for said given image part of interest in dependence upon said prioritized set of candidate depth values.
10. An image display apparatus, comprising:
a motion estimation device as claimed in claim 8 for furnishing motion data;
an image signal processing device for processing an image signal to obtain an enhanced image signal in dependence upon said motion data; and
a display device for displaying the enhanced image signal.