System and method for the detection of de-interlacing of scaled video
AssigneeNXP B.V.
InventorsDmitry Znamenskiy, Claus Nico Cordes
Life of this patent
Estimating the term…
Abstract
A system, apparatus, and method are provided for a video detector that computes a measure of how much a given video content resembles one of a de-interlaced video content or a progressive video content. More particularly, the present invention determines the position of original and interpolated lines and the scaling factor of an input content whenever that content was scaled after de-interlacing.
Claims26
A method for detection of de-interlacing, comprising:
- providing a set of lines (y,n)∈ S and a discriminating function D(y,n) for the provided set, where n is a specific frame number and y is a line number within the specified frame;
- defining an odd grid as a subset of pixels to consist of lines (y,n)∈ S with line number y in frames with frame number n where sum y+n is odd and defining an even grid to consist of lines (y,n)∈ S with line number y in frames with frame number n where sum y+n is even;
- evaluating a pre-set discriminating function D(y,n) for each of the lines (y,n)∈ S;
- computing an average A1 of the evaluated discriminating function D(y,n) on a subset of pixels belonging to the even grid and average A2 of the evaluated discriminating function D(y,n) on a subset of pixels belonging to odd grid;
- computing a de-interlacing quality measure based on the difference between A1 and A2;
- and
- classifying the even grid as the grid of original lines if A1>A2, otherwise classifying the odd grid as the grid of original lines.
An apparatus for detection of de-interlacing, comprising:
- a module that performs a pre-set discriminating function D(y,n) on a subset of the set of lines (y,n)∈ S where n is a specific frame number and y is a line number within the specified frame,
- a grid determination module that defines an odd grid as a subset of pixels consisting of lines (y,n)∈ S with line number y in frames with frame number n where sum y+n is odd and defining an even grid as a subset of pixels consisting of lines (y,n)∈ S with line number y in frames with frame number n where sum y+n is even;
- an averaging module that computes an average A1 by executing the discriminating function D(y,n) on the pixels belonging to the even grid and average A2 by executing the discriminating function D(y,n) on a the pixels belonging to the odd grid;
- a comparison module that outputs a de-interlacing quality measure based on a comparison between the values A1 and A2;
- and
- a classification module that classifies the even grid as the grid of original lines if A1>A2, and otherwise classifies the odd grid as the grid of original lines.
A method for detecting de-interlacing for scaled video, comprising:
- for every scaling ratio α of a given set α ∈ I computing a measure, ρ(α) which indicates the visibility of de-interlacing artifacts given that the content was scaled with vertical factor α, by performing the following:
- a. evaluating a pre-set discriminating function Dα(y,n) for a given set of lines (y,n)∈ S, and
- b. convolving the discriminating function Dα(y,n) with a pre-set parity function which uses α as a scaling coefficient;
- analyzing the measure ρ(α) and defining a de-interlacing quality measure Q that is inversely proportional to visibility of peak(s) in μ(α) as a function of α;
- analyzing the measure ρ(α) and defining the most probable vertical scaling ratio β as α corresponding to the maximum of ρ(α);
- and
- defining the positions corresponding to original lines in the scaled content as maxima of
- cos(π(βy+n)+φ(β).
The method of claim 16, wherein where the pre-set discriminating function D(y,n) is based on a weighted sum of different specific discriminating functions.
An apparatus for detection of de-interlacing for scaled video having a vertical scaling factor α, comprising:
- a de-interlacing artifact visibility detection module comprising sub modules:
- a. a discrimination function module that performs a pre-set discriminating function Dα(y,n) on a subset of the set of lines (y,n)∈ S where n is a specific frame number and y is a line number within the specified frame, and
- b. a convolving discriminating function module that convolves the discriminating function Dα(y,n) using a pre-set parity function which uses α as a scaling coefficient to obtain ρ(α) which indicates the visibility of de-interlacing artifacts;
- a de-interlacing quality definition and analysis module that analyzes the measure ρ(α) and defines a de-interlacing quality measure Q that is inversely proportional to visibility of peak(s) in ρ(α) as a function of α, defines the most probable vertical scaling ration β as α corresponding to the maximum of ρ(α) and defines the positions corresponding to original lines in the scaled content as maxima of cos(π(βy+n)+φ(β)).
Dependencies are inferred from the claim text.