kdu_line_buf

Instances of this structure manage the buffering of a single line of sample values, whether image samples or subband samples. For the reversible path, samples must be absolute 32- or 16-bit integers. For irreversible processing, samples have a normalized representation, where the nominal range of the data is typically -0.5 to +0.5 (actual nominal ranges for the data supplied to a kdu_encoder or kdu_analysis object or retrieved from a kdu_decoder or kdu_synthesis object may be explicitly set in the relevant constructor). These normalized quantities may have either a true 32-bit floating point representation or a 16-bit fixed-point representation. In the latter case, the least significant KDU_FIX_POINT bits of the integer are interpreted as binary fraction bits, so that 2^{KDU_FIX_POINT} represents a normalized value of 1.0.

For simplicity, the object always allocates space to allow access to 2 samples before the first nominal sample and 8 samples after the last nominal sample in the line. This should prove sufficient to accommodate the needs of boundary extension schemes and efficient SIMD (multi-word) implementations of the sample data transformations.

kdu_line_buf {Java: Kdu_line_buf}

pre_create {Java: Pre_create}

create {Java: Create}

destroy {Java: Destroy}

deactivate {Java: Deactivate}

activate {Java: Activate}

is_active {Java: Is_active}

operator!

get_buf32

get_buf16

get_floats {Java: Get_floats}

set_floats {Java: Set_floats}

get_ints (2 forms) {Java: Get_ints}

set_ints (2 forms) {Java: Set_ints}

get_width {Java: Get_width}

is_absolute {Java: Is_absolute}