| 摘要 |
A method for encoding a data value array comprises: generating, from input data values, respective complementary most-significant and least-significant data portions, the most-significant data portion representing plural most significant bits (MSBs) of that value, and the least-significant data portion dependent upon the remaining least significant bits (LSBs); generating one or more data sets (e.g. a significance map) indicating positions of the most-significant data portions of predetermined magnitude ranges; binary encoding the data sets; and including data defining least-significant portions in the output data stream. Also claimed is a similar method for encoding a data array as data sets and escape codes comprising unary and non-unary coded portions - for values not included in the data sets; a coding parameter between 0 and an upper limit is set for the non-unary coded portion and an offset is added to define minimum least significant data portion size. Also disclosed is an image data encoding method comprising frequency-transforming input data, generating a frequency-transformed coefficient array by matrix multiplication, according to a maximum dynamic range and using transform matrices having a data precision; dynamic range and data precision are selected according to image data bit depth. |
