Intra-prediction estimation using approximate reconstructed samples
AssigneeTexas Instruments Incorporated
InventorsRanga Ramanujam Srinivasan, Mahant Siddaramanna, Naveen Srinivasamurthy
Life of this patent
Estimating the term…
Abstract
A method for intra-prediction estimation is provided that includes determining a best intra-prediction mode for a block of samples, wherein at least some of the neighboring samples used for intra-prediction estimation include approximate reconstructed samples, applying approximate reconstruction to the block of samples using the best intra-prediction mode to generate a block of approximate reconstructed samples, and storing the block of approximate reconstructed samples for use in intra-prediction estimation of other blocks of samples.
Claims20
A method for approximate reconstruction comprising:
- selecting, by at least one processor, an intra-prediction mode;
- applying, by the at least one processor, the intra-prediction mode to an original block of video data to generate a predicted block;
- subtracting, by the at least one processor, the predicted block from the original block to generate a residual block;
- applying, by the at least one processor, a simplified transform to the residual block to generate transform coefficients, wherein the simplified transform is based on a discrete cosine transform (DCT) corresponding to multiplication of coefficients of DCT basis vectors by a constant;
- applying, by the at least one processor, a simplified quantization to the transform coefficients to generate quantized transform coefficients;
- applying, by the at least one processor, an inverse quantization to the quantized transform coefficients to generate dequantized coefficients;
- applying, by the at least one processor, an inverse simplified transform to the dequantized coefficients to generate a block of reconstructed residuals;
- adding, by the at least one processor, the predicted block to the block of reconstructed residuals to generate a block of approximate reconstructed samples;
- and
- storing, in a memory, the block of approximate reconstructed samples.
A device comprising:
- at least one processor;
- and
- a non-transitory computer readable storage medium storing a program for execution by the at least one processor, the program including instructions to cause the at least one processor to:
- select an intra-prediction mode;
- apply the intra-prediction mode to an original block of video data to generate a predicted block;
- subtract the predicted block from the original block to generate a residual block;
- apply a simplified transform to the residual block to generate transform coefficients, wherein the simplified transform is based on a discrete cosine transform (DCT) corresponding to multiplication of coefficients of DCT basis vectors by a constant;
- apply a simplified quantization to the transform coefficients to generate quantized transform coefficients;
- apply an inverse quantization to the quantized transform coefficients to generate dequantized coefficients;
- apply an inverse simplified transform to the dequantized coefficients to generate a block of reconstructed residuals;
- add the predicted block to the block of reconstructed residuals to generate a block of approximate reconstructed samples;
- and
- store the block of approximate reconstructed samples in a memory.
A non-transitory computer readable medium comprising instructions which, when executed, cause processor circuitry to at least:
- apply a intra-prediction mode to an original block of video data to generate a predicted block;
- subtract the predicted block from the original block to generate a residual block;
- apply a simplified transform to the residual block to generate transform coefficients by multiplying coefficients of vectors by a constant to generate products and applying a floor function to the products;
- apply a simplified quantization to the transform coefficients to generate quantized transform coefficients;
- apply an inverse quantization to the quantized transform coefficients to generate dequantized coefficients;
- apply an inverse simplified transform to the dequantized coefficients to generate a block of reconstructed residuals;
- and
- add the predicted block to the block of reconstructed residuals to generate a block of approximate reconstructed samples.
Dependencies are inferred from the claim text.