Decoding of predicted DC coefficient without division
View Patent ↗A technique for eliminating the division in decoding a predicted DC coefficient includes calculating and storing the values of 1//DC_scalar in a table. This allows one to convert a division into a multiplication.
1. A processing circuit comprising:
means for calculating all possible values for an inverted variable used in decoding predicted zero-frequency transform coefficients; and
means for storing all the values in a lookup table;
means for decoding the predicted zero-frequency transform coefficients with the values stored in the lookup table without using a divide unit.
2. The processing circuit of claim 1
where the means for calculating all possible values includes means for calculating an equivalence of all possible values for an inverted scalar value of a quantization parameter for a macro block calculated using rounding division; and
where the means for storing all the values in a lookup table comprises means for storing a value exactly equal to the inverse of the scalar value of a quantization parameter for a macro block for a predetermined level of precision.
3. The processing circuit of claim 1
where the means for calculating all possible values comprises means for calculating for each possible value of a scalar value of a quantization parameter for a macro block a value exactly equal to an inverse of the scalar value of quantization parameter for a macro block for a predetermined level of precision.
4. The processing circuit of claim 1
where the means for calculating all possible values comprises means for calculating, for a predetermined precision level, the zero-frequency coefficient of a macro block used for prediction purposes times the inverted variable used in decoding predicted zero-frequency transform coefficients for all values of the scalar value of a quantization parameter for a macro block.
5. The processing circuit of claim 1
where the means for calculating all possible values comprises means for calculating thirty-nine possible values for the inverted scalar value of a quantization parameter for a macro block; and
where the means for storing all the values in a lookup table comprises means for storing a value exactly equal to the inverse scalar of a quantization parameter for a macro block for a precision being used.
6. A processing circuit comprising:
means for storing all thirty-nine values of an inverted variable used in decoding a zero-frequency coefficient in a lookup table; and
means for calculating the zero-frequency coefficient using the lookup table to avoid using rounding division.
7. A processing circuit comprising:
means for storing all thirty-nine values of an inverted variable used in decoding zero-frequency coefficients calculated once off-line in a lookup table; and
means for calculating the zero-frequency coefficients without accessing a divide unit by repeatedly accessing the stored lookup table.
8. A system comprising:
encoded images;
a storage circuit to store a lookup table; and
a processor to decode a bit stream in the encoded images in software with said lookup table where a divide unit is not used when decoding predicted zero-frequency transform coefficients.