Method of filtering a video sequence image from spurious motion effects

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
• distance
• =
• ∑
• i
• =
• -
• N
• 2
• N
• 2
• -
• 1
• ⁢
• ∑
• j
• =
• -
• M
• 2
• M
• 2
• -
• 1
• ⁢
• f
• a
• ⁡
• (
• i
• ,
• j
• )
• ·
• d
• ⁡
• (
• i
• ,
• j
• )
• ∑
• i
• =
• -
• N
• 2
• N
• 2
• -
• 1
• ⁢
• ∑
• j
• =
• -
• M
• 2
• M
• 2
• -
• 1
• ⁢
• f
• a
• ⁡
• (
• i
• ,
• j
• )
• -
• f
• a
• ⁡
• (
• 0
• ,
• 0
• )
• where d(i,j) is a Euclidean distance from an origin of the I and j axes and fa(i,j) is a Fourier spectrum component defined in the interval [−N/2, N/2−1]×[−M/2, M/2−1] of a sequence of size N×M,
• a second parameter density according to the following formula
• density
• =
• ∑
• k
• =
• 1
• componentsNumber
• ⁢
• neighbors
• ⁡
• (
• k
• )
• componentsNumber
• where componentsNumber represents a number of non-zero values of the Fourier spectrum and neighbors(k) a number of non-zero values close to the component k;
• discriminating the current block of pixels as either a substantially regular or a substantially irregular pattern by a fuzzy logic decision rule using as antecedents said first and second parameters;
• processing the substantially regular blocks of pixels with an edge preserving noise filtering algorithm or with an edge sharpening algorithm, generating corresponding enhanced blocks of pixels;
• generating said enhanced output image using said enhanced blocks of pixels.

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.

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.