Method of filtering a video sequence image from spurious motion effects
First Claim
1. A method, comprising:
- filtering, using one or more processors, a current image of an input video sequence from spurious motion effects, the filtering including;
calculating motion vectors for blocks of pixels of the current image of the video sequence with a block matching algorithm carried out on corresponding blocks of pixels of the current image and of a preceding image that precedes the current image in the sequence;
selecting a subset of said motion vectors by deselecting;
motion vectors calculated from blocks of the current image dissimilar from corresponding blocks of the preceding image,motion vectors strongly different from motion vectors of surrounding blocks, andmotion vectors associated with pixels of homogeneous areas of the current image;
calculating roto-translational and zooming parameters describing spurious global motion between the current image and the preceding image, by processing the motion vectors of said subset through the following recursive procedure;
calculating, for each current motion vector of said subset, a corresponding expected motion vector estimated in function of roto-translational and zooming parameters relative to the preceding image,calculating for each current motion vector of said subset a respective error value in function of the motion vector of the corresponding block of the preceding image and the corresponding expected motion vector,comparing said error values with a first threshold and storing in a memory the current motion vectors of said subset having error values that are smaller than said first threshold and deleting from said memory motion vectors previously stored therein having error values that are larger than said first threshold, andcalculating said roto-translational and zooming parameters for the current image in function of motion vectors stored in said memory and of current motion vectors of said subset; and
generating a filtered output image from said current image of the input video sequence by compensating spurious motion effects described by said roto-translational and zooming parameters.
1 Assignment
0 Petitions
Accused Products
Abstract
According to the novel method, roto-translational and zooming parameters describing spurious motion effects are determined by exploiting any of the many block matching algorithms commonly used for motion estimation for calculating a motion vector for all or for a selected number of blocks of pixels of the current frame that is processed. Some of the so calculated motion vectors are not taken into account for estimating spurious motion effects. The roto-translational and zooming parameters describing what is considered to be spurious global motion between a current frame and the precedent frame of the sequence, are calculated by processing the selected motion vectors of blocks of pixels of the frame through a recursive procedure that includes computing error values and readjusting the roto-translational and zooming parameters based on the error values.
-
Citations
28 Claims
-
1. A method, comprising:
-
filtering, using one or more processors, a current image of an input video sequence from spurious motion effects, the filtering including; calculating motion vectors for blocks of pixels of the current image of the video sequence with a block matching algorithm carried out on corresponding blocks of pixels of the current image and of a preceding image that precedes the current image in the sequence; selecting a subset of said motion vectors by deselecting; motion vectors calculated from blocks of the current image dissimilar from corresponding blocks of the preceding image, motion vectors strongly different from motion vectors of surrounding blocks, and motion vectors associated with pixels of homogeneous areas of the current image; calculating roto-translational and zooming parameters describing spurious global motion between the current image and the preceding image, by processing the motion vectors of said subset through the following recursive procedure; calculating, for each current motion vector of said subset, a corresponding expected motion vector estimated in function of roto-translational and zooming parameters relative to the preceding image, calculating for each current motion vector of said subset a respective error value in function of the motion vector of the corresponding block of the preceding image and the corresponding expected motion vector, comparing said error values with a first threshold and storing in a memory the current motion vectors of said subset having error values that are smaller than said first threshold and deleting from said memory motion vectors previously stored therein having error values that are larger than said first threshold, and calculating said roto-translational and zooming parameters for the current image in function of motion vectors stored in said memory and of current motion vectors of said subset; and generating a filtered output image from said current image of the input video sequence by compensating spurious motion effects described by said roto-translational and zooming parameters. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18)
-
-
19. A method, comprising:
-
generating an enhanced output image by enhancing a digital input image using one or more processors, the enhancing including; subdividing the input image into blocks of pixels; discriminating substantially regular from substantially irregular blocks of pixels by calculating, for each current block of pixels; a first parameter distance according to the following formula - View Dependent Claims (20, 21)
-
-
22. A non-transitory computer-readable medium comprising contents that cause a computing device to implement a method that includes:
-
filtering a current image of an input video sequence from spurious motion effects, the filtering including; calculating motion vectors for blocks of pixels of the current image of the video sequence with a block matching algorithm carried out on corresponding blocks of pixels of the current image and of a preceding image that precedes the current image in the sequence; selecting a subset of said motion vectors by deselecting; motion vectors calculated from blocks of the current image dissimilar from corresponding blocks of the preceding image, motion vectors strongly different from motion vectors of surrounding blocks, and motion vectors associated with pixels of homogeneous areas of the current image; calculating roto-translational and zooming parameters describing spurious global motion between the current image and the preceding image, by processing the motion vectors of said subset through the following recursive procedure; calculating, for each current motion vector of said subset, a corresponding expected motion vector estimated in function of roto-translational and zooming parameters relative to the preceding image, calculating for each current motion vector of said subset a respective error value in function of the motion vector of the corresponding block of the preceding image and the corresponding expected motion vector, comparing said error values with a first threshold and storing in a memory the current motion vectors of said subset having error values that are smaller than said first threshold and deleting from said memory motion vectors previously stored therein having error values that are larger than said first threshold, and calculating said roto-translational and zooming parameters for the current image in function of motion vectors stored in said memory and of current motion vectors of said subset; and generating a filtered output image from said current image of the input video sequence by compensating spurious motion effects described by said roto-translational and zooming parameters. - View Dependent Claims (23, 24, 25)
-
-
26. A device, comprising:
-
filtering means for filtering a current image of an input video sequence from spurious motion effects, the filtering including; calculating motion vectors for blocks of pixels of the current image of the video sequence with a block matching algorithm carried out on corresponding blocks of pixels of the current image and of a preceding image that precedes the current image in the sequence; selecting a subset of said motion vectors by deselecting; motion vectors calculated from blocks of the current image dissimilar from corresponding blocks of the preceding image, motion vectors strongly different from motion vectors of surrounding blocks, and motion vectors associated with pixels of homogeneous areas of the current image; calculating roto-translational and zooming parameters describing spurious global motion between the current image and the preceding image, by processing the motion vectors of said subset through the following recursive procedure; calculating, for each current motion vector of said subset, a corresponding expected motion vector estimated in function of roto-translational and zooming parameters relative to the preceding image, calculating for each current motion vector of said subset a respective error value in function of the motion vector of the corresponding block of the preceding image and the corresponding expected motion vector, comparing said error values with a first threshold and storing in a memory the current motion vectors of said subset having error values that are smaller than said first threshold and deleting from said memory motion vectors previously stored therein having error values that are larger than said first threshold, and calculating said roto-translational and zooming parameters for the current image in function of motion vectors stored in said memory and of current motion vectors of said subset; and an image generator structured to generate a filtered output image from said current image of the input video sequence by compensating spurious motion effects described by said roto-translational and zooming parameters. - View Dependent Claims (27, 28)
-
Specification