/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// Intel License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2000, Intel Corporation, all rights reserved.
// Third party copyrights are property of their respective owners.
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// * The name of Intel Corporation may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#include "_cxcore.h"
/****************************************************************************************\
* Mean and StdDev calculation *
\****************************************************************************************/
#define ICV_MEAN_SDV_COI_CASE( worktype, sqsumtype, \
sqr_macro, len, cn ) \
for( ; x <= (len) - 4*(cn); x += 4*(cn))\
{ \
worktype t0 = src[x]; \
worktype t1 = src[x + (cn)]; \
\
s0 += t0 + t1; \
sq0 += (sqsumtype)(sqr_macro(t0)) + \
(sqsumtype)(sqr_macro(t1)); \
\
t0 = src[x + 2*(cn)]; \
t1 = src[x + 3*(cn)]; \
\
s0 += t0 + t1; \
sq0 += (sqsumtype)(sqr_macro(t0)) + \
(sqsumtype)(sqr_macro(t1)); \
} \
\
for( ; x < (len); x += (cn) ) \
{ \
worktype t0 = src[x]; \
\
s0 += t0; \
sq0 += (sqsumtype)(sqr_macro(t0)); \
}
#define ICV_MEAN_SDV_CASE_C1( worktype, sqsumtype, sqr_macro, len ) \
ICV_MEAN_SDV_COI_CASE( worktype, sqsumtype, sqr_macro, len, 1 )
#define ICV_MEAN_SDV_CASE_C2( worktype, sqsumtype, \
sqr_macro, len ) \
for( ; x < (len); x += 2 ) \
{ \
worktype t0 = (src)[x]; \
worktype t1 = (src)[x + 1]; \
\
s0 += t0; \
sq0 += (sqsumtype)(sqr_macro(t0)); \
s1 += t1; \
sq1 += (sqsumtype)(sqr_macro(t1)); \
}
#define ICV_MEAN_SDV_CASE_C3( worktype, sqsumtype, \
sqr_macro, len ) \
for( ; x < (len); x += 3 ) \
{ \
worktype t0 = (src)[x]; \
worktype t1 = (src)[x + 1]; \
worktype t2 = (src)[x + 2]; \
\
s0 += t0; \
sq0 += (sqsumtype)(sqr_macro(t0)); \
s1 += t1; \
sq1 += (sqsumtype)(sqr_macro(t1)); \
s2 += t2; \
sq2 += (sqsumtype)(sqr_macro(t2)); \
}
#define ICV_MEAN_SDV_CASE_C4( worktype, sqsumtype, \
sqr_macro, len ) \
for( ; x < (len); x += 4 ) \
{ \
worktype t0 = (src)[x]; \
worktype t1 = (src)[x + 1]; \
\
s0 += t0; \
sq0 += (sqsumtype)(sqr_macro(t0)); \
s1 += t1; \
sq1 += (sqsumtype)(sqr_macro(t1)); \
\
t0 = (src)[x + 2]; \
t1 = (src)[x + 3]; \
\
s2 += t0; \
sq2 += (sqsumtype)(sqr_macro(t0)); \
s3 += t1; \
sq3 += (sqsumtype)(sqr_macro(t1)); \
}
#define ICV_MEAN_SDV_MASK_COI_CASE( worktype, sqsumtype, \
sqr_macro, len, cn ) \
for( ; x <= (len) - 4; x += 4 ) \
{ \
worktype t0; \
if( mask[x] ) \
{ \
t0 = src[x*(cn)]; pix++; \
s0 += t0; \
sq0 += sqsumtype(sqr_macro(t0)); \
} \
\
if( mask[x+1] ) \
{ \
t0 = src[(x+1)*(cn)]; pix++; \
s0 += t0; \
sq0 += sqsumtype(sqr_macro(t0)); \
} \
\
if( mask[x+2] ) \
{ \
t0 = src[(x+2)*(cn)]; pix++; \
s0 += t0; \
sq0 += sqsumtype(sqr_macro(t0)); \
} \
\
if( mask[x+3] ) \
{ \
t0 = src[(x+3)*(cn)]; pix++; \
s0 += t0; \
sq0 += sqsumtype(sqr_macro(t0)); \
} \
} \
\
for( ; x < (len); x++ ) \
{ \
if( mask[x] ) \
{ \
worktype t0 = src[x*(cn)]; pix++; \
s0 += t0; \
sq0 += sqsumtype(sqr_macro(t0)); \
} \
}
#define ICV_MEAN_SDV_MASK_CASE_C1( worktype, sqsumtype, sqr_macro, len ) \
ICV_MEAN_SDV_MASK_COI_CASE( worktype, sqsumtype, sqr_macro, len, 1 )
#define ICV_MEAN_SDV_MASK_CASE_C2( worktype, sqsumtype,\
sqr_macro, len ) \
for( ; x < (len); x++ ) \
{ \
if( mask[x] ) \
{ \
worktype t0 = src[x*2]; \
worktype t1 = src[x*2+1]; \
pix++; \
s0 += t0; \
sq0 += sqsumtype(sqr_macro(t0)); \
s1 += t1; \
sq1 += sqsumtype(sqr_macro(t1)); \
} \
}
#define ICV_MEAN_SDV_MASK_CASE_C3( worktype, sqsumtype,\
sqr_macro, len ) \
for( ; x < (len); x++ ) \
{ \
if( mask[x] ) \
{ \
worktype t0 = src[x*3]; \
worktype t1 = src[x*3+1]; \
worktype t2 = src[x*3+2]; \
pix++; \
s0 += t0; \
sq0 += sqsumtype(sqr_macro(t0)); \
s1 += t1; \
sq1 += sqsumtype(sqr_macro(t1)); \
s2 += t2; \
sq2 += sqsumtype(sqr_macro(t2)); \
} \
}
#define ICV_MEAN_SDV_MASK_CASE_C4( worktype, sqsumtype,\
sqr_macro, len ) \
for( ; x < (len); x++ ) \
{ \
if( mask[x] ) \
{ \
worktype t0 = src[x*4]; \
worktype t1 = src[x*4+1]; \
pix++; \
s0 += t0; \
sq0 += sqsumtype(sqr_macro(t0)); \
s1 += t1; \
sq1 += sqsumtype(sqr_macro(t1)); \
t0 = src[x*4+2]; \
t1 = src[x*4+3]; \
s2 += t0; \
sq2 += sqsumtype(sqr_macro(t0)); \
s3 += t1; \
sq3 += sqsumtype(sqr_macro(t1)); \
} \
}
////////////////////////////////////// entry macros //////////////////////////////////////
#define ICV_MEAN_SDV_ENTRY_COMMON() \
int pix; \
double scale, tmp; \
step /= sizeof(src[0])
#define ICV_MEAN_SDV_ENTRY_C1( sumtype, sqsumtype ) \
sumtype s0 = 0; \
sqsumtype sq0 = 0; \
ICV_MEAN_SDV_ENTRY_COMMON()
#define ICV_MEAN_SDV_ENTRY_C2( sumtype, sqsumtype ) \
sumtype s0 = 0, s1 = 0; \
sqsumtype sq0 = 0, sq1 = 0; \
ICV_MEAN_SDV_ENTRY_COMMON()
#define ICV_MEAN_SDV_ENTRY_C3( sumtype, sqsumtype ) \
sumtype s0 = 0, s1 = 0, s2 = 0; \
sqsumtype sq0 = 0, sq1 = 0, sq2 = 0; \
ICV_MEAN_SDV_ENTRY_COMMON()
#define ICV_MEAN_SDV_ENTRY_C4( sumtype, sqsumtype ) \
sumtype s0 = 0, s1 = 0, s2 = 0, s3 = 0; \
sqsumtype sq0 = 0, sq1 = 0, sq2 = 0, sq3 = 0; \
ICV_MEAN_SDV_ENTRY_COMMON()
#define ICV_MEAN_SDV_ENTRY_BLOCK_COMMON( block_size ) \
int remaining = block_size; \
ICV_MEAN_SDV_ENTRY_COMMON()
#define ICV_MEAN_SDV_ENTRY_BLOCK_C1( sumtype, sqsumtype, \
worktype, sqworktype, block_size ) \
sumtype sum0 = 0; \
sqsumtype sqsum0 = 0; \
worktype s0 = 0; \
sqworktype sq0 = 0; \
ICV_MEAN_SDV_ENTRY_BLOCK_COMMON( block_size )
#define ICV_MEAN_SDV_ENTRY_BLOCK_C2( sumtype, sqsumtype, \
worktype, sqworktype, block_size ) \
sumtype sum0 = 0, sum1 = 0; \
sqsumtype sqsum0 = 0, sqsum1 = 0; \
worktype s0 = 0, s1 = 0; \
sqworktype sq0 = 0, sq1 = 0; \
ICV_MEAN_SDV_ENTRY_BLOCK_COMMON( block_size )
#define ICV_MEAN_SDV_ENTRY_BLOCK_C3( sumtype, sqsumtype, \
worktype, sqworktype, block_size ) \
sumtype sum0 = 0, sum1 = 0, sum2 = 0; \
sqsumtype sqsum0 = 0, sqsum1 = 0, sqsum2 = 0; \
worktype s0 = 0, s1 = 0, s2 = 0; \
sqworktype sq0 = 0, sq1 = 0, sq2 = 0; \
ICV_MEAN_SDV_ENTRY_BLOCK_COMMON( block_size )
#define ICV_MEAN_SDV_ENTRY_BLOCK_C4( sumtype, sqsumtype, \
worktype, sqworktype, block_size ) \
sumtype sum0 = 0, sum1 = 0, sum2 = 0, sum3 = 0; \
sqsumtype sqsum0 = 0, sqsum1 = 0, sqsum2 = 0, sqsum3 = 0; \
worktype s0 = 0, s1 = 0, s2 = 0, s3 = 0; \
sqworktype sq0 = 0, sq1 = 0, sq2 = 0, sq3 = 0; \
ICV_MEAN_SDV_ENTRY_BLOCK_COMMON( block_size )
/////////////////////////////////////// exit macros //////////////////////////////////////
#define ICV_MEAN_SDV_EXIT_COMMON() \
scale = pix ? 1./pix : 0
#define ICV_MEAN_SDV_EXIT_CN( total, sqtotal, idx ) \
ICV_MEAN_SDV_EXIT_COMMON(); \
mean[idx] = tmp = scale*(double)total##idx; \
tmp = scale*(double)sqtotal##idx - tmp*tmp; \
sdv[idx] = sqrt(MAX(tmp,0.))
#define ICV_MEAN_SDV_EXIT_C1( total, sqtotal ) \
ICV_MEAN_SDV_EXIT_COMMON(); \
ICV_MEAN_SDV_EXIT_CN( total, sqtotal, 0 )
#define ICV_MEAN_SDV_EXIT_C2( total, sqtotal ) \
ICV_MEAN_SDV_EXIT_COMMON(); \
ICV_MEAN_SDV_EXIT_CN( total, sqtotal, 0 ); \
ICV_MEAN_SDV_EXIT_CN( total, sqtotal, 1 )
#define ICV_MEAN_SDV_EXIT_C3( total, sqtotal ) \
ICV_MEAN_SDV_EXIT_COMMON(); \
ICV_MEAN_SDV_EXIT_CN( total, sqtotal, 0 ); \
ICV_MEAN_SDV_EXIT_CN( total, sqtotal, 1 ); \
ICV_MEAN_SDV_EXIT_CN( total, sqtotal, 2 )
#define ICV_MEAN_SDV_EXIT_C4( total, sqtotal ) \
ICV_MEAN_SDV_EXIT_COMMON(); \
ICV_MEAN_SDV_EXIT_CN( total, sqtotal, 0 ); \
ICV_MEAN_SDV_EXIT_CN( total, sqtotal, 1 ); \
ICV_MEAN_SDV_EXIT_CN( total, sqtotal, 2 ); \
ICV_MEAN_SDV_EXIT_CN( total, sqtotal, 3 )
////////////////////////////////////// update macros /////////////////////////////////////
#define ICV_MEAN_SDV_UPDATE_COMMON( block_size )\
remaining = block_size
#define ICV_MEAN_SDV_UPDATE_C1( block_size ) \
ICV_MEAN_SDV_UPDATE_COMMON( block_size ); \
sum0 += s0; sqsum0 += sq0; \
s0 = 0; sq0 = 0
#define ICV_MEAN_SDV_UPDATE_C2( block_size ) \
ICV_MEAN_SDV_UPDATE_COMMON( block_size ); \
sum0 += s0; sqsum0 += sq0; \
sum1 += s1; sqsum1 += sq1; \
s0 = s1 = 0; sq0 = sq1 = 0
#define ICV_MEAN_SDV_UPDATE_C3( block_size ) \
ICV_MEAN_SDV_UPDATE_COMMON( block_size ); \
sum0 += s0; sqsum0 += sq0; \
sum1 += s1; sqsum1 += sq1; \
sum2 += s2; sqsum2 += sq2; \
s0 = s1 = s2 = 0; sq0 = sq1 = sq2 = 0
#define ICV_MEAN_SDV_UPDATE_C4( block_size ) \
ICV_MEAN_SDV_UPDATE_COMMON( block_size ); \
sum0 += s0; sqsum0 += sq0; \
sum1 += s1; sqsum1 += sq1; \
sum2 += s2; sqsum2 += sq2; \
sum3 += s3; sqsum3 += sq3; \
s0 = s1 = s2 = s3 = 0; sq0 = sq1 = sq2 = sq3 = 0
#define ICV_DEF_MEAN_SDV_BLOCK_FUNC_2D( flavor, cn, arrtype, \
sumtype, sqsumtype, worktype, \
sqworktype, block_size, sqr_macro ) \
IPCVAPI_IMPL( CvStatus, icvMean_StdDev_##flavor##_C##cn##R, \
( const arrtype* src, int step, \
CvSize size, double* mean, double* sdv ), \
(src, step, size, mean, sdv) ) \
{ \
ICV_MEAN_SDV_ENTRY_BLOCK_C##cn( sumtype, sqsumtype, \
worktype, sqworktype, (block_size)*(cn) ); \
pix = size.width * size.height; \
size.width *= (cn); \
\
for( ; size.height--; src += step ) \
{ \
int x = 0; \
while( x < size.width ) \
{ \
int limit = MIN( remaining, size.width - x ); \
remaining -= limit; \
limit += x; \
ICV_MEAN_SDV_CASE_C##cn( worktype, sqworktype, \
sqr_macro, limit ); \
if( remaining == 0 ) \
{ \
ICV_MEAN_SDV_UPDATE_C##cn( (block_size)*(cn) ); \
} \
} \
} \
\
ICV_MEAN_SDV_UPDATE_C##cn(0); \
ICV_MEAN_SDV_EXIT_C##cn( sum, sqsum ); \
return CV_OK; \
}
#define ICV_DEF_MEAN_SDV_FUNC_2D( flavor, cn, arrtype, \
sumtype, sqsumtype, worktype ) \
IPCVAPI_IMPL( CvStatus, icvMean_StdDev_##flavor##_C##cn##R, \
( const arrtype* src, int step, \
CvSize size, double* mean, double* sdv ), \
(src, step, size, mean, sdv) ) \
{ \
ICV_MEAN_SDV_ENTRY_C##cn( sumtype, sqsumtype ); \
pix = size.width * size.height; \
size.width *= (cn); \
\
for( ; size.height--; src += step ) \
{ \
int x = 0; \
ICV_MEAN_SDV_CASE_C##cn( worktype, sqsumtype, \
CV_SQR, size.width ); \
} \
\
ICV_MEAN_SDV_EXIT_C##cn( s, sq ); \
return CV_OK; \
}
#define ICV_DEF_MEAN_SDV_BLOCK_FUNC_2D_COI( flavor, arrtype, \
sumtype, sqsumtype, worktype, \
sqworktype, block_size, sqr_macro ) \
static CvStatus CV_STDCALL icvMean_StdDev_##flavor##_CnCR \
( const arrtype* src, int step, \
CvSize size, int cn, int coi, \
double* mean, double* sdv ) \
{ \
ICV_MEAN_SDV_ENTRY_BLOCK_C1( sumtype, sqsumtype, \
worktype, sqworktype, (block_size)*(cn) ); \
pix = size.width * size.height; \
size.width *= (cn); \
src += coi - 1; \
\
for( ; size.height--; src += step ) \
{ \
int x = 0; \
while( x < size.width ) \
{ \
int limit = MIN( remaining, size.width - x ); \
remaining -= limit; \
limit += x; \
ICV_MEAN_SDV_COI_CASE( worktype, sqworktype, \
sqr_macro, limit, cn); \
if( remaining == 0 ) \
{ \
ICV_MEAN_SDV_UPDATE_C1( (block_size)*(cn) ); \
} \
} \
} \
\
ICV_MEAN_SDV_UPDATE_C1(0); \
ICV_MEAN_SDV_EXIT_C1( sum, sqsum ); \
return CV_OK; \
}
#define ICV_DEF_MEAN_SDV_FUNC_2D_COI( flavor, arrtype, \
sumtype, sqsumtype, worktype )\
static CvStatus CV_STDCALL icvMean_StdDev_##flavor##_CnCR \
( const arrtype* src, int step, CvSize size,\
int cn, int coi, double* mean, double* sdv )\
{ \
ICV_MEAN_SDV_ENTRY_C1( sumtype, sqsumtype ); \
pix = size.width * size.height; \
size.width *= (cn); \
src += coi - 1; \
\
for( ; size.height--; src += step ) \
{ \
int x = 0; \
ICV_MEAN_SDV_COI_CASE( worktype, sqsumtype, \
CV_SQR, size.width, cn ); \
} \
\
ICV_MEAN_SDV_EXIT_C1( s, sq ); \
return CV_OK; \
}
#define ICV_DEF_MEAN_SDV_MASK_BLOCK_FUNC_2D( flavor, cn, \
arrtype, sumtype, sqsumtype, worktype, \
sqworktype, block_size, sqr_macro ) \
IPCVAPI_IMPL( CvStatus, icvMean_StdDev_##flavor##_C##cn##MR, \
( const arrtype* src, int step, \
const uchar* mask, int maskstep, \
CvSize size, double* mean, double* sdv ), \
(src, step, mask, maskstep, size, mean, sdv))\
{ \
ICV_MEAN_SDV_ENTRY_BLOCK_C##cn( sumtype, sqsumtype, \
worktype, sqworktype, block_size ); \
pix = 0; \
\
for( ; size.height--; src += step, mask += maskstep ) \
{ \
int x = 0; \
while( x < size.width ) \
{ \
int limit = MIN( remaining, size.width - x ); \
remaining -= limit; \
limit += x; \
ICV_MEAN_SDV_MASK_CASE_C##cn( worktype, sqworktype, \
sqr_macro, limit ); \
if( remaining == 0 ) \
{ \
ICV_MEAN_SDV_UPDATE_C##cn( block_size ); \
} \
} \
} \
\
ICV_MEAN_SDV_UPDATE_C##cn(0); \
ICV_MEAN_SDV_EXIT_C##cn( sum, sqsum ); \
return CV_OK; \
}
#define ICV_DEF_MEAN_SDV_MASK_FUNC_2D( flavor, cn, arrtype, \
sumtype, sqsumtype, worktype)\
IPCVAPI_IMPL( CvStatus, icvMean_StdDev_##flavor##_C##cn##MR, \
( const arrtype* src, int step, \
const uchar* mask, int maskstep, \
CvSize size, double* mean, double* sdv ), \
(src, step, mask, maskstep, size, mean, sdv))\
{ \
ICV_MEAN_SDV_ENTRY_C##cn( sumtype, sqsumtype ); \
pix = 0; \
\
for( ; size.height--; src += step, mask += maskstep ) \
{ \
int x = 0; \
ICV_MEAN_SDV_MASK_CASE_C##cn( worktype, sqsumtype, \
CV_SQR, size.width ); \
} \
\
ICV_MEAN_SDV_EXIT_C##cn( s, sq ); \
return CV_OK; \
}
#define ICV_DEF_MEAN_SDV_MASK_BLOCK_FUNC_2D_COI( flavor, \
arrtype, sumtype, sqsumtype, worktype, \
sqworktype, block_size, sqr_macro ) \
static CvStatus CV_STDCALL icvMean_StdDev_##flavor##_CnCMR \
( const arrtype* src, int step, \
const uchar* mask, int maskstep, \
CvSize size, int cn, int coi, \
double* mean, double* sdv ) \
{ \
ICV_MEAN_SDV_ENTRY_BLOCK_C1( sumtype, sqsumtype, \
worktype, sqworktype, block_size ); \
pix = 0; \
src += coi - 1; \
\
for( ; size.height--; src += step, mask += maskstep ) \
{ \
int x = 0; \
while( x < size.width ) \
{ \
int limit = MIN( remaining, size.width - x ); \
remaining -= limit; \
limit += x; \
ICV_MEAN_SDV_MASK_COI_CASE( worktype, sqworktype, \
sqr_macro, limit, cn ); \
if( remaining == 0 ) \
{ \
ICV_MEAN_SDV_UPDATE_C1( block_size ); \
} \
} \
} \
\
ICV_MEAN_SDV_UPDATE_C1(0); \
ICV_MEAN_SDV_EXIT_C1( sum, sqsum ); \
return CV_OK; \
}
#define ICV_DEF_MEAN_SDV_MASK_FUNC_2D_COI( flavor, arrtype, \
sumtype, sqsumtype, worktype ) \
static CvStatus CV_STDCALL icvMean_StdDev_##flavor##_CnCMR \
( const arrtype* src, int step, \
const uchar* mask, int maskstep, \
CvSize size, int cn, int coi, \
double* mean, double* sdv ) \
{ \
ICV_MEAN_SDV_ENTRY_C1( sumtype, sqsumtype ); \
pix = 0; \
src += coi - 1; \
\
for( ; size.height--; src += step, mask += maskstep ) \
{ \
int x = 0; \
ICV_MEAN_SDV_MASK_COI_CASE( worktype, sqsumtype, \
CV_SQR, size.width, cn ); \
} \
\
ICV_MEAN_SDV_EXIT_C1( s, sq ); \
return CV_OK; \
}
#define ICV_DEF_MEAN_SDV_BLOCK_ALL( flavor, arrtype, sumtype, sqsumtype,\
worktype, sqworktype, block_size, sqr_macro)\
ICV_DEF_MEAN_SDV_BLOCK_FUNC_2D( flavor, 1, arrtype, sumtype, sqsumtype, \
worktype, sqworktype, block_size, sqr_macro)\
ICV_DEF_MEAN_SDV_BLOCK_FUNC_2D( flavor, 2, arrtype, sumtype, sqsumtype, \
worktype, sqworktype, block_size, sqr_macro)\
ICV_DEF_MEAN_SDV_BLOCK_FUNC_2D( flavor, 3, arrtype, sumtype, sqsumtype, \
worktype, sqworktype, block_size, sqr_macro)\
ICV_DEF_MEAN_SDV_BLOCK_FUNC_2D( flavor, 4, arrtype, sumtype, sqsumtype, \
worktype, sqworktype, block_size, sqr_macro)\
ICV_DEF_MEAN_SDV_BLOCK_FUNC_2D_COI( flavor, arrtype, sumtype, sqsumtype,\
worktype, sqworktype, block_size, sqr_macro)\
\
ICV_DEF_MEAN_SDV_MASK_BLOCK_FUNC_2D( flavor, 1, arrtype, sumtype, \
sqsumtype, worktype, sqworktype, block_size, sqr_macro ) \
ICV_DEF_MEAN_SDV_MASK_BLOCK_FUNC_2D( flavor, 2, arrtype, sumtype, \
sqsumtype, worktype, sqworktype, block_size, sqr_macro ) \
ICV_DEF_MEAN_SDV_MASK_BLOCK_FUNC_2D( flavor, 3, arrtype, sumtype, \
sqsumtype, worktype, sqworktype, block_size, sqr_macro ) \
ICV_DEF_MEAN_SDV_MASK_BLOCK_FUNC_2D( flavor, 4, arrtype, sumtype, \
sqsumtype, worktype, sqworktype, block_size, sqr_macro ) \
ICV_DEF_MEAN_SDV_MASK_BLOCK_FUNC_2D_COI( flavor, arrtype, sumtype, \
sqsumtype, worktype, sqworktype, block_size, sqr_macro )
#define ICV_DEF_MEAN_SDV_ALL( flavor, arrtype, sumtype, sqsumtype, worktype ) \
ICV_DEF_MEAN_SDV_FUNC_2D( flavor, 1, arrtype, sumtype, sqsumtype, worktype ) \
ICV_DEF_MEAN_SDV_FUNC_2D( flavor, 2, arrtype, sumtype, sqsumtype, worktype ) \
ICV_DEF_MEAN_SDV_FUNC_2D( flavor, 3, arrtype, sumtype, sqsumtype, worktype ) \
ICV_DEF_MEAN_SDV_FUNC_2D( flavor, 4, arrtype, sumtype, sqsumtype, worktype ) \
ICV_DEF_MEAN_SDV_FUNC_2D_COI( flavor, arrtype, sumtype, sqsumtype, worktype ) \
\
ICV_DEF_MEAN_SDV_MASK_FUNC_2D(flavor, 1, arrtype, sumtype, sqsumtype, worktype) \
ICV_DEF_MEAN_SDV_MASK_FUNC_2D(flavor, 2, arrtype, sumtype, sqsumtype, worktype) \
ICV_DEF_MEAN_SDV_MASK_FUNC_2D(flavor, 3, arrtype, sumtype, sqsumtype, worktype) \
ICV_DEF_MEAN_SDV_MASK_FUNC_2D(flavor, 4, arrtype, sumtype, sqsumtype, worktype) \
ICV_DEF_MEAN_SDV_MASK_FUNC_2D_COI( flavor, arrtype, sumtype, sqsumtype, worktype )
ICV_DEF_MEAN_SDV_BLOCK_ALL( 8u, uchar, int64, int64, unsigned, unsigned, 1 << 16, CV_SQR_8U )
ICV_DEF_MEAN_SDV_BLOCK_ALL( 16u, ushort, int64, int64, unsigned, int64, 1 << 16, CV_SQR )
ICV_DEF_MEAN_SDV_BLOCK_ALL( 16s, short, int64, int64, int, int64, 1 << 16, CV_SQR )
ICV_DEF_MEAN_SDV_ALL( 32s, int, double, double, double )
ICV_DEF_MEAN_SDV_ALL( 32f, float, double, double, double )
ICV_DEF_MEAN_SDV_ALL( 64f, double, double, double, double )
#define icvMean_StdDev_8s_C1R 0
#define icvMean_StdDev_8s_C2R 0
#define icvMean_StdDev_8s_C3R 0
#define icvMean_StdDev_8s_C4R 0
#define icvMean_StdDev_8s_CnCR 0
#define icvMean_StdDev_8s_C1MR 0
#define icvMean_StdDev_8s_C2MR 0
#define icvMean_StdDev_8s_C3MR 0
#define icvMean_StdDev_8s_C4MR 0
#define icvMean_StdDev_8s_CnCMR 0
CV_DEF_INIT_BIG_FUNC_TAB_2D( Mean_StdDev, R )
CV_DEF_INIT_FUNC_TAB_2D( Mean_StdDev, CnCR )
CV_DEF_INIT_BIG_FUNC_TAB_2D( Mean_StdDev, MR )
CV_DEF_INIT_FUNC_TAB_2D( Mean_StdDev, CnCMR )
CV_IMPL void
cvAvgSdv( const CvArr* img, CvScalar* _mean, CvScalar* _sdv, const void* mask )
{
CvScalar mean = {{0,0,0,0}};
CvScalar sdv = {{0,0,0,0}};
static CvBigFuncTable meansdv_tab;
static CvFuncTable meansdvcoi_tab;
static CvBigFuncTable meansdvmask_tab;
static CvFuncTable meansdvmaskcoi_tab;
static int inittab = 0;
CV_FUNCNAME("cvMean_StdDev");
__BEGIN__;
int type, coi = 0;
int mat_step, mask_step = 0;
CvSize size;
CvMat stub, maskstub, *mat = (CvMat*)img, *matmask = (CvMat*)mask;
if( !inittab )
{
icvInitMean_StdDevRTable( &meansdv_tab );
icvInitMean_StdDevCnCRTable( &meansdvcoi_tab );
icvInitMean_StdDevMRTable( &meansdvmask_tab );
icvInitMean_StdDevCnCMRTable( &meansdvmaskcoi_tab );
inittab = 1;
}
if( !CV_IS_MAT(mat) )
CV_CALL( mat = cvGetMat( mat, &stub, &coi ));
type = CV_MAT_TYPE( mat->type );
if( CV_MAT_CN(type) > 4 && coi == 0 )
CV_ERROR( CV_StsOutOfRange, "The input array must have at most 4 channels unless COI is set" );
size = cvGetMatSize( mat );
mat_step = mat->step;
if( !mask )
{
if( CV_IS_MAT_CONT( mat->type ))
{
size.width *= size.height;
size.height = 1;
mat_step = CV_STUB_STEP;
}
if( CV_MAT_CN(type) == 1 || coi == 0 )
{
CvFunc2D_1A2P func = (CvFunc2D_1A2P)(meansdv_tab.fn_2d[type]);
if( !func )
CV_ERROR( CV_StsBadArg, cvUnsupportedFormat );
IPPI_CALL( func( mat->data.ptr, mat_step, size, mean.val, sdv.val ));
}
else
{
CvFunc2DnC_1A2P func = (CvFunc2DnC_1A2P)
(meansdvcoi_tab.fn_2d[CV_MAT_DEPTH(type)]);
if( !func )
CV_ERROR( CV_StsBadArg, cvUnsupportedFormat );
IPPI_CALL( func( mat->data.ptr, mat_step, size,
CV_MAT_CN(type), coi, mean.val, sdv.val ));
}
}
else
{
CV_CALL( matmask = cvGetMat( matmask, &maskstub ));
mask_step = matmask->step;
if( !CV_IS_MASK_ARR( matmask ))
CV_ERROR( CV_StsBadMask, "" );
if( !CV_ARE_SIZES_EQ( mat, matmask ))
CV_ERROR( CV_StsUnmatchedSizes, "" );
if( CV_IS_MAT_CONT( mat->type & matmask->type ))
{
size.width *= size.height;
size.height = 1;
mat_step = mask_step = CV_STUB_STEP;
}
if( CV_MAT_CN(type) == 1 || coi == 0 )
{
CvFunc2D_2A2P func = (CvFunc2D_2A2P)(meansdvmask_tab.fn_2d[type]);
if( !func )
CV_ERROR( CV_StsBadArg, cvUnsupportedFormat );
IPPI_CALL( func( mat->data.ptr, mat_step, matmask->data.ptr,
mask_step, size, mean.val, sdv.val ));
}
else
{
CvFunc2DnC_2A2P func = (CvFunc2DnC_2A2P)
(meansdvmaskcoi_tab.fn_2d[CV_MAT_DEPTH(type)]);
if( !func )
CV_ERROR( CV_StsBadArg, cvUnsupportedFormat );
IPPI_CALL( func( mat->data.ptr, mat_step,
matmask->data.ptr, mask_step,
size, CV_MAT_CN(type), coi, mean.val, sdv.val ));
}
}
__END__;
if( _mean )
*_mean = mean;
if( _sdv )
*_sdv = sdv;
}
/* End of file */