METHOD FOR CARRYING OUT INTEGER APPROXIMATION OF TRANSFORM COEFFICIENTS, AND CODER AND DECODER

摘要

For obtaining cosine transform coefficients that have undergone an integer approximation, a restricted set of values is specified. The coefficients of the base vectors for the submatrices are selected, while satisfying the orthogonality condition, in such a manner that the sum of its squares results in the square of the dc component coefficient. These coefficients are used to derive the coefficients of the ac components. The inventive method has the advantage that a uniform standardization and quantization factor can be used for all coefficients for the purpose of quantization and standardization.