IMG_corr_gen_i16s_c16s


Detailed Description


Functions

void IMG_corr_gen_i16s_c16s (short *restrict x, short *restrict h, short *restrict y, int m, int pitch)


Function Documentation

void IMG_corr_gen_i16s_c16s ( short *restrict  x,
short *restrict  h,
short *restrict  y,
int  m,
int  pitch 
)

Description:
This routine performs a generalized correlation with a 1 by M tap filter. It can be called repetitively to form an arbitrary MxN 2D generalized correlation kernel. The correlation sum is stored as half words. The input pixel and mask coefficient are assumed to be in signed 16-bit format. No restrictions apply to 'pitch' and 'M'.
Parameters:
x Pointer to an input image of 16-bit pixels
h Pointer to a 16-bit linear (1xM) filter mask
y Output correlation sum(s)
m Length of linear filter (taps)
pitch Number of columns in input image
Algorithm:
The natural C implementation has no restrictions. The optimized intrinsic C code has restrictions as noted in Assumptions below.
Assumptions:
  • The input array and output array should not overlap
  • The input array must be 32-bit aligned
  • The output array must be 64-bit aligned
  • The mask array must be 16-bit aligned
  • The internal accuracy is 16 bits. The convloution sum should not exceed 16 bits (signed).
  • The output array size must have at least (pitch - M + 8) elements
  • The pitch parameter must be greater than the filter length (M)
Implementation Notes:
  • This code is fully interruptible
  • This code is compatible with C66x processors
Benchmarks:
See IMGLIB_Test_Report.html for cycle and memory information.


Copyright 2012, Texas Instruments Incorporated