SPRUJ28E November 2021 – September 2024 AM68 , AM68A , TDA4AL-Q1 , TDA4VE-Q1 , TDA4VL-Q1
In YCbCr mode, the offset table defines a (x,y) vector for a regular grid of output points. The grid can be fully sampled or down sampled. A fully sampled grid will define an offset vector for every output pixel, defining exactly where to fetch the input data to compute the output pixel. This is the most precise definition and can capture rapidly changing offset tables. The drawback is that it will require a large amount of memory bandwidth as the LDC engine will be reading offset values for every output pixel. Since most offset tables are not expected to change rapidly in a small spatial region, LDC supports reading a subsampled offset table. Offset tables can be subsampled by powers of two in both horizontal and vertical directions and the subsampling factor is set in the register, VPAC_LDC_MESHTABLE_CFG[2-0] M. This mode conserves memory bandwidth by reducing the amount of data read to describe the offset vectors, but requires more hardware to interpolate the missing offset vectors. LDC supports bilinear interpolation to interpolate missing offset vectors.
The mapping procedure is described by the following series of equations. Given an output pixel at location, we compute the input pixel location.
Clip new locations to previously fetched mesh block boundaries.
Graphically, this procedure is shown in Figure 6-128.
The input coordinate (xi,yi ) refers to final input frame co-ordinates and is used to fetch the appropriate input pixels for interpolation. The interpolation procedure is described in . Before fetching data(xi,yi ), is clipped to the input frame boundary.