Method of filtering pixels in a video encoding process
View Patent ↗A method and apparatus, particularly suited to SIMD instruction sets, to filter streaming video information encoded under a predictive encoding algorithm specified under video encoding standards, such as MPEG 4 or H.264/AVC. The filtering operation de-blocks or removes unwanted borders in the perceived video. During the filtering process, a series of filtering mask is generated based on temporal and spatial statistics of predictive encoded video information, which is then recursively applied to the video in order to gate filtered or unfiltered video to an output channel according to coefficients of the masks. The filtering mask effectively yields a decision or rule-based map that transforms the video on a pixel-by-pixel basis thereby avoiding complex and processor-intensive decision tree logic customarily required to process individual pixels of successive macroblocks that may have different filtering requirements.
1. A method of filtering predictive encoded streaming video information in order to filter individual elements of pixel groups, said method comprising:
obtaining statistical parameters from video information encoded under a predictive encoding algorithm,
utilizing a single-instruction-multiple-data instruction, generating a set of filtering masks based on said statistical parameters, and
in response to a threshold value derived from said statistical parameters, selectively utilizing said filtering mask to filter individual elements of said pixel groups by gating one of filtered and unfiltered pixel groups over a video channel,
wherein the generating step comprises generating filtering masks including a mask M=(|P 1 −Q 1 |<A)AND(|P 1 −P 2 |<B)AND(|Q 1 −Q 2 |<B), a mask M p =(|P 1 −P 3 |<B)AND(M) and mask M q =(|Q 1 −Q 3 |<B)AND(M) where P x and Q x , x being said specified integers, are pixel vectors having multiple pixel elements, A and B represent respective vectors of threshold values, and AND represent logical “AND” operation.
2. The method of claim 1 , wherein said encoding algorithm utilizes a deviation threshold that determines a maximum deviation between filtered and unfiltered elements, said method further comprising:
modifying the deviation threshold utilized during predictive encoding according to said filtering mask.
3. The method of claim 2 , wherein said modifying comprises utilizing a modified deviation threshold C 0 ′ determined by subtracting masks M p and M q from C 0 , where C 0 is a vector representation of threshold value, C 0 ′ is a vector representation of modified threshold values, and mask M p =(|P 1 −P 3 |<B)AND(M) and mask M q =(|Q 1 −Q 3 |<B)AND(M).