Performs affine warp transformation from input image (8-bit) to output image (8-bit) using bilinear interpolation. No border handling in done in this function, so it is expected that appropriate padding is done to the input buffer.
- Method:
- The affine warp kernel is computed by using the following equation:
x0 = x * warpMatrix[0] + y * warpMatrix[2] + warpMatrix[4];
y0 = x * warpMatrix[1] + y * warpMatrix[3] + warpMatrix[5];
dst(x,y) = BilinearInterpolate(src(x0, y0))
- Parameters
-
| [in] | src[] | Pointer to array containing first input image (UQ8.0) |
| [in] | src_addr[] | Pointer to structure containing dimensional information of src |
| [out] | dst[] | Pointer to array containing output image (UQ8.0) |
| [in] | dst_addr[] | Pointer to structure containing dimensional information of dst |
| [in] | warpMatrix[] | Pointer to array containing the affine coefficients (F32) |
| [in] | srcOffsetX | Parameter indicating x offset of src pointer relative to start of valid data in broader image (SQ15.0) |
| [in] | srcOffsetY | Parameter indicating y offset of src pointer relative to start of valid data in broader image (SQ15.0) |
| [in] | dstOffsetX | Parameter indicating x offset of dst pointer relative to start of output valid data in broader image (SQ15.0) |
| [in] | dstOffsetY | Parameter indicating y offset of dst pointer relative to start of output valid data in broader image (SQ15.0) |
- Assumptions:
- I/O buffer pointers are assumed to be not aliased.
- If a user wants to divide processing of the image into smaller blocks, then it can use the dstOffsetX and dstOffsetY to indicate the starting coordinate of the output block relative to the start of the full output image, and srcOffsetX and srcOffsetY to indicate the starting coordinate of the input block relative to the start of the full input image.
- Be sure to allocate padding around the border (or fetch enough pad pixels from the input for interior edges when breaking into blocks), for the warp, or else the function may fetch from an out-of-bounds memory location. For each dimension, the required fetch amount depends on the warpMatrix coefficients.
- If padding has been applied to the start of the input image, then the src pointer should point to the start of the image including the padding, and the srcOffsetX and srcOffsetY will be negative numbers (e.g. If 2 rows and 2 columns of padding have been added, then srcOffsetX and srcOffsetY will both be -2).
