This section describes functions that perform Discrete Cosine Transform (DCT) plus some additional operations - level shift, quantization (or dequantization) - and are specific for JPEG codec. These functions are listed in Table “Functions that combine Quantization, DCT, and Level Shift”.
| Function Base Name | Description |
|---|---|
| DCTQuantFwd8x8_JPEG | Performs forward DCT and quantization. |
| DCTQuantFwd8x8LS_JPEG | Performs forward DCT, quantization and level shift. |
| DCTQuantInv8x8_JPEG | Performs inverse DCT and dequantization. |
| DCTQuantInv8x8LS_JPEG | Performs inverse DCT, dequantization and level shift. |
| DCTQuantInv8x8LS_1x1_JPEG, CTQuantInv8x8LS_2x2_JPEG, DCTQuantInv8x8LS_4x4_JPEG | Performs fast inverse DCT, dequantization and level shift for 8x8 block with zero high frequency components. |
| DCTQuantInv8x8To4x4LS_JPEG, DCTQuantInv8x8To2x2LS_JPEG | Performs fast inverse DCT, dequantization, level shift and downsampling. |
These functions implement component operations in accordance with their definition by [ISO10918], specifically:
- for DCT, as in [ISO10918], Annex A.3.3, FDCT and IDCT;
- for quantization/dequantization, as in [ISO10918], Annex A.3.4, DCT Coefficient Quantization and Dequantization;
- for level shift, as in [ISO10918], Annex A.3.1, Level Shift.
Note
The IPP functions described in this section use quantization tables in special format that is suitable for fast encoding/decoding. Initial raw quantization tables must be previously transformed using functions ippiQuantFwdTableInit_JPEG and ippiQuantInvTableInit_JPEG.