摘要:
This disclosure describes techniques for mitigating rounding errors in a fixed-point transform associated with video coding by applying a variable localized bit-depth increase at the transform. More specifically, the techniques include selecting a constant value based on a size of a fixed-point transform in a video coding device and applying a variable localized bit-depth increase at the transform with a value equal to the constant value. Applying the variable localized bit-depth increase includes left-shifting a transform input signal by a number of bits equal to the constant value before the fixed-point transform, and right-shifting a transform output signal by a number of bits equal to the constant value after the fixed-point transform. The constant value is selected from a plurality of constant values stored on the video coding device. Each of the constant values is pre-calculated for one of a plurality of different transform sizes supported by the video coding.
摘要:
This disclosure describes techniques for mitigating rounding errors in a fixed-point transform associated with video coding by applying a variable localized bit-depth increase at the transform. More specifically, the techniques include selecting a constant value based on a size of a fixed-point transform in a video coding device and applying a variable localized bit-depth increase at the transform with a value equal to the constant value. Applying the variable localized bit-depth increase includes left-shifting a transform input signal by a number of bits equal to the constant value before the fixed-point transform, and right-shifting a transform output signal by a number of bits equal to the constant value after the fixed-point transform. The constant value is selected from a plurality of constant values stored on the video coding device. Each of the constant values is pre-calculated for one of a plurality of different transform sizes supported by the video coding.
摘要:
During the prediction stage of a video encoding and/or decoding process, a video coder can use relatively longer filters for certain motion vectors pointing to certain sub-pixel positions and relatively shorter filters for motion vectors pointing to other sub-pixel positions, where a longer filter generally refers to an interpolation filter with a greater number of filter coefficients, also called taps, while a shorter filter generally refers to an interpolation filter with fewer taps.
摘要:
During the prediction stage of a video encoding and/or decoding process, a video coder can use relatively longer filters for certain motion vectors pointing to certain sub-pixel positions and relatively shorter filters for motion vectors pointing to other sub-pixel positions, where a longer filter generally refers to an interpolation filter with a greater number of filter coefficients, also called taps, while a shorter filter generally refers to an interpolation filter with fewer taps.
摘要:
A video encoder may encode video data by adaptively selecting between one-eighth-pixel and one-quarter-pixel precision motion vectors, and signal the selected precision. In one example, an apparatus includes a video encoder to encode a block of video data using a one-eighth-pixel precision motion vector when use of the one-eighth-pixel precision motion vector is determined to be preferable for the block over a one-quarter-pixel precision motion vector, and to generate a signal value indicative of the use of the one-eighth-pixel precision motion vector for the block, and an output interface to output the encoded block and the signal value. A video decoder may be configured to receive the signal value and the encoded block, analyze the signal value to determine whether the block was encoded using one-eighth-pixel precision or one-quarter-pixel precision, and decode the block based on the determination.
摘要:
This disclosure relates to techniques for reducing a cost of coding prediction information in video coding. Video blocks in a generalized P/B (GPB) frame are encoded using up to two motion vectors calculated from reference pictures in two separate reference picture lists that are identical. Video blocks of a GPB frame may, therefore, be encoded using a bidirectional prediction mode with a first motion vector from a reference picture in a first reference picture list and a second motion vector from the same or substantially similar reference picture in a second reference picture list. The techniques include jointly coding the first and second motion vectors for a video block of a GPB frame. The techniques include coding the first motion vector relative to a first motion predictor generated from a motion vector of a neighboring block, and coding the second motion vector relative to the first motion vector.
摘要:
A video encoder may encode video data by adaptively selecting between one-eighth-pixel and one-quarter-pixel precision motion vectors, and signal the selected precision. In one example, an apparatus includes a video encoder to encode a block of video data using a one-eighth-pixel precision motion vector when use of the one-eighth-pixel precision motion vector is determined to be preferable for the block over a one-quarter-pixel precision motion vector, and to generate a signal value indicative of the use of the one-eighth-pixel precision motion vector for the block, and an output interface to output the encoded block and the signal value. A video decoder may be configured to receive the signal value and the encoded block, analyze the signal value to determine whether the block was encoded using one-eighth-pixel precision or one-quarter-pixel precision, and decode the block based on the determination.
摘要:
A method of coding video data includes determining a candidate motion vector for each of one or more candidate portions of a video frame and determining a current motion vector for a current portion of a current frame. The current motion vector identifies a portion of a reference frame that at least partially matches the current portion of the current frame. The method also includes calculating a motion vector difference between the current motion vector and each of the candidate motion vectors, selecting one of the candidate motion vectors based on the calculated motion vector differences, signaling an index identifying the candidate portion having the selected one of the candidate motion vectors, and signaling the corresponding motion vector difference calculated with respect to the selected one of the candidate motion vectors.
摘要:
The example techniques described in this disclosure provide for an efficient manner to encode or decode a video block of a picture using a single reference picture list. The single reference picture list may include identifiers for reference picture or pictures used to encode or decode the video block. In some examples, a video encoder or decoder may encode or decode a video block that is predicted from two reference pictures using the single reference picture list, and encode or decode a video block that is predicted from one reference picture using the same, single reference picture list.
摘要:
A video block syntax element indicates whether all of the partitions of a video block are predicted based on a same reference list and no greater than quarter-pixel accuracy is used. If the video block syntax element is set, partition-level signaling of the reference lists is avoided. If the video block syntax element is not set, partition-level signaling of the reference lists occurs. If the video block syntax element is set, partition-level syntax elements may be used for each of the partitions of the video block, wherein the partition-level syntax elements each identify one of the reference lists and motion vector accuracy for a given one of the partitions.