Method for reduced bit-depth quantization

1. A method for reconstruction of a compressed image signal by dequantization and inverse transformation of an image difference signal, said method comprising:

receiving a plurality of quantized coefficient levels that define an image;

receiving a quantization parameter (QP);

determining a periodic mantissa portion matrix depending on QP % P using a constant parameter (P);

determining an exponential portion depending on QP/P;

computing reconstructed transform coefficients from said plurality of quantized coefficient levels, said periodic mantissa portion and said exponential portion;

performing an inverse transformation on said reconstructed transform coefficients thereby yielding scaled samples;

computing reconstructed samples, by normalizing the scaled samples using a constant normalization factor (M).

2. The method of claim 1 wherein P=6 and M=6 and wherein said computing a reconstructed transform coefficient comprises using a formula equivalent to {tilde over (w)} α =[c α ·R α QP % P ]<<(QP/P) and said computing reconstructed samples comprises using a formula equivalent to x″ α =({tilde over (x)}′ α +(1<<(M−1)))>>M.

3. A method as described in claim 1 wherein said constant parameter (P) is equal to 6.

4. A method as described in claim 1 wherein said constant normalization factor (M) is equal to 6.

5. A method as described in claim 1 wherein said constant parameter (P) is equal to 6 and said constant normalization factor (M) is equal to 6.

6. A method for reconstruction of a compressed image signal by dequantization and inverse transformation of an image difference signal, said method comprising:

receiving a plurality of quantized coefficient levels (c α ) that define an image;

receiving a quantization parameter (QP);

determining a mantissa portion matrix, R α QP % P ;

determining an exponential portion, depending on QP/P;

computing reconstructed transform coefficients ({tilde over (w)} α ) wherein {tilde over (w)} α =[c α ·R α QP % P ]<<(QP/P);

performing an inverse transformation on said reconstructed transform coefficients ({tilde over (w)} α ) thereby yielding transformed samples ({tilde over (x)}′ α );

computing reconstructed samples, (x″ α ) using a formula equivalent to {tilde over (x)}″ α =({tilde over (x)}′ α +(1<<(M−1)))>>M, wherein M is a constant normalization factor.

7. An apparatus for reconstruction of a compressed image signal by dequantization and inverse transformation of an image difference signal, said apparatus comprising:

a coefficient receiver for receiving a plurality of quantized coefficient levels that define an image;

a quantization parameter receiver for receiving a quantization parameter (QP);

a mantissa portion processor for determining a periodic mantissa portion matrix depending on QP % P using a constant parameter (P);

an exponential portion processor for determining an exponential portion depending on QP/P;

a transform coefficient processor for computing reconstructed transform coefficients from said plurality of quantized coefficient levels, said periodic mantissa portion and said exponential portion;

a transformation processor for performing an inverse transformation on said reconstructed transform coefficients thereby yielding scaled samples;

a reconstructed sample processor for computing reconstructed samples, by normalizing the scaled samples using a constant normalization factor (M).