another similar improvment

This commit is contained in:
Desmis 2019-07-01 10:17:07 +02:00
parent b371629b23
commit dc4788e00b
2 changed files with 24 additions and 10 deletions

View File

@ -4009,8 +4009,9 @@ void ImProcFunctions::fftw_convol_blur(float *input, float *output, int bfw, int
** radius = sigma for kernel
** n_x n_y relative width and high for kernel
** Gaussian blur is given by G(x,y) = (1/2*PI*sigma) * exp(-(x2 + y2) / 2* sigma2)
** its traduction in Fourier transform is G(x,y) = exp((-sigma)*(PI * x2 + PI * y2))
** after several test the only result that works very well is with fftkern = 0 and algo = 0
** its traduction in Fourier transform is G(x,y) = exp((-sigma)*(PI * x2 + PI * y2)), for some authors it is not sigma but sigma^2..I have tried...huge diffrences with Gaussianblur
** after several test the only result that works very well is with fftkern = 0 and algo = 0, and as there is big differences with Gaussianblur, I put an empirical correction in Ipretinex and Iplocalcontrast
** in fact no importance....if it is this function (for sigma) or another... we are not in research :)
*/
BENCHFUN

View File

@ -962,19 +962,32 @@ void ImProcFunctions::MSRLocal(int sp, bool fftw, int lum, LabImage * bufreti, L
}
float *buffer = new float[W_L * H_L];
float kr = 1.f;
float kr = 1.f;//on FFTW
float kg = 1.f;//on Gaussianblur
for (int scale = scal - 1; scale >= 0; scale--) {
// printf("retscale=%f scale=%i \n", mulradiusfftw * RetinexScales[scale], scale);
//emprical adjustement between FFTW radius and Gaussainblur
//under 50 ==> 10.f
//above 400 ==> 1.f
//emprical adjustement between FFTW radius and Gaussainblur
//under 50 ==> 10.f
// 400 ==> 1.f
float sigm = RetinexScales[scale];
float mulradiusfftw = 1.f;
float ak = -9.f / 350.f;
float bk = 10.f - 50.f * ak;
kr = ak * sigm + bk;
if(sigm < 50.f) kr = 10.f;
if(sigm > 400.f) kr = 1.f;
//above 400 at 5000 ==> 20.f
if(sigm > 400.f) {//increase ==> 5000
float ka = 19.f / 4600.f;
float kb = 1.f - 400 * ka;
kr = ka * sigm + kb;
float kga = -0.5f / 4600.f;
float kgb = 1.f - 400.f * kga;
kg = kga * sigm + kgb;
if(sigm > 5000.f) {
kr = ka * 5000.f + kb;
kg = kga * 5000.f + kgb;
}
}
if(!fftw) {
#ifdef _OPENMP
@ -984,11 +997,11 @@ void ImProcFunctions::MSRLocal(int sp, bool fftw, int lum, LabImage * bufreti, L
if (scale == scal - 1)
{
gaussianBlur(src, out, W_L, H_L, RetinexScales[scale], buffer);
gaussianBlur(src, out, W_L, H_L, kg * RetinexScales[scale], buffer);
} else // reuse result of last iteration
{
// out was modified in last iteration => restore it
gaussianBlur(out, out, W_L, H_L, sqrtf(SQR(RetinexScales[scale]) - SQR(RetinexScales[scale + 1])), buffer);
gaussianBlur(out, out, W_L, H_L, sqrtf(SQR(kg * RetinexScales[scale]) - SQR(kg * RetinexScales[scale + 1])), buffer);
}
}
} else {