eddy/fitsmaniplib
1
0
Fork 0
mirror of https://github.com/eddyem/fitsmaniplib.git synced 2025-12-06 02:35:13 +03:00
Code Issues Packages Projects Releases Wiki Activity
fitsmaniplib/median.c

699 lines
22 KiB
C
Raw Blame History

/*
* This file is part of the FITSmaniplib project.
* Copyright 2019 Edward V. Emelianov <edward.emelianoff@gmail.com>, <eddy@sao.ru>.
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
// FOR MEDIATOR:
// Copyright (c) 2011 ashelly.myopenid.com under <http://www.opensource.org/licenses/mit-license>
// FOR opt_medXX:
// Copyright (c) 1998 Nicolas Devillard. Public domain.
// FOR qickselect:
// "Numerical recipes in C", Second Edition,
// Cambridge University Press, 1992, Section 8.5, ISBN 0-521-43108-5
// Code by Nicolas Devillard - 1998. Public domain.
// TODO: resolve problem with borders
#include "FITSmanip.h"
#include "local.h"
// largest radius for adaptive median filter
#define LARGEST_ADPMED_RADIUS (3)
#define ELEM_SWAP(a, b) {register double t = a; a = b; b = t;}
#define PIX_SORT(a, b) {if (p[a] > p[b]) ELEM_SWAP(p[a], p[b]);}
/*
* simplest short functions for median calculation
*/
// even values are from "FAST, EFFICIENT MEDIAN FILTERS WITH EVEN LENGTH WINDOWS",
// J.P. HAVLICEK, K.A. SAKADY, G.R.KATZ
static double opt_med2(double *p){
return (p[0] + p[1]) * 0.5;
}
static double opt_med3(double *p){
PIX_SORT(0, 1); PIX_SORT(1, 2); PIX_SORT(0, 1);
return (p[1]);
}
static double opt_med4(double *p){
PIX_SORT(0, 2); PIX_SORT(1, 3);
PIX_SORT(0, 1); PIX_SORT(2, 3);
return (p[1] + p[2]) / 2.;
}
static double opt_med5(double *p){
PIX_SORT(0, 1); PIX_SORT(3, 4); PIX_SORT(0, 3);
PIX_SORT(1, 4); PIX_SORT(1, 2); PIX_SORT(2, 3) ;
PIX_SORT(1, 2);
return (p[2]);
}
static double opt_med6(double *p){
PIX_SORT(1, 2); PIX_SORT(3, 4);
PIX_SORT(0, 1); PIX_SORT(2, 3); PIX_SORT(4, 5);
PIX_SORT(1, 2); PIX_SORT(3, 4);
PIX_SORT(0, 1); PIX_SORT(2, 3); PIX_SORT(4, 5);
PIX_SORT(1, 2); PIX_SORT(3, 4);
return (p[2] + p[3]) / 2.;
}
static double opt_med7(double *p){
PIX_SORT(0, 5); PIX_SORT(0, 3); PIX_SORT(1, 6);
PIX_SORT(2, 4); PIX_SORT(0, 1); PIX_SORT(3, 5);
PIX_SORT(2, 6); PIX_SORT(2, 3); PIX_SORT(3, 6);
PIX_SORT(4, 5); PIX_SORT(1, 4); PIX_SORT(1, 3);
PIX_SORT(3, 4);
return (p[3]);
}
// optimal Batcher's sort for 8 elements (http://myopen.googlecode.com/svn/trunk/gtkclient_tdt/include/fast_median.h)
static double opt_med8(double *p){
PIX_SORT(0, 4); PIX_SORT(1, 5); PIX_SORT(2, 6);
PIX_SORT(3, 7); PIX_SORT(0, 2); PIX_SORT(1, 3);
PIX_SORT(4, 6); PIX_SORT(5, 7); PIX_SORT(2, 4);
PIX_SORT(3, 5); PIX_SORT(0, 1); PIX_SORT(2, 3);
PIX_SORT(4, 5); PIX_SORT(6, 7); PIX_SORT(1, 4);
PIX_SORT(3, 6);
return (p[3] + p[4]) / 2.;
}
static double opt_med9(double *p){
PIX_SORT(1, 2); PIX_SORT(4, 5); PIX_SORT(7, 8);
PIX_SORT(0, 1); PIX_SORT(3, 4); PIX_SORT(6, 7);
PIX_SORT(1, 2); PIX_SORT(4, 5); PIX_SORT(7, 8);
PIX_SORT(0, 3); PIX_SORT(5, 8); PIX_SORT(4, 7);
PIX_SORT(3, 6); PIX_SORT(1, 4); PIX_SORT(2, 5);
PIX_SORT(4, 7); PIX_SORT(4, 2); PIX_SORT(6, 4);
PIX_SORT(4, 2);
return (p[4]);
}
static double opt_med16(double *p){
PIX_SORT(0, 8); PIX_SORT(1, 9); PIX_SORT(2, 10); PIX_SORT(3, 11);
PIX_SORT(4, 12); PIX_SORT(5, 13); PIX_SORT(6, 14); PIX_SORT(7, 15);
PIX_SORT(0, 4); PIX_SORT(1, 5); PIX_SORT(2, 6); PIX_SORT(3, 7);
PIX_SORT(8, 12); PIX_SORT(9, 13); PIX_SORT(10, 14); PIX_SORT(11, 15);
PIX_SORT(4, 8); PIX_SORT(5, 9); PIX_SORT(6, 10); PIX_SORT(7, 11);
PIX_SORT(0, 2); PIX_SORT(1, 3); PIX_SORT(4, 6); PIX_SORT(5, 7);
PIX_SORT(8, 10); PIX_SORT(9, 11); PIX_SORT(12, 14); PIX_SORT(13, 15);
PIX_SORT(2, 8); PIX_SORT(3, 9); PIX_SORT(6, 12); PIX_SORT(7, 13);
PIX_SORT(2, 4); PIX_SORT(3, 5); PIX_SORT(6, 8); PIX_SORT(7, 9);
PIX_SORT(10, 12); PIX_SORT(11, 13); PIX_SORT(0, 1); PIX_SORT(2, 3);
PIX_SORT(4, 5); PIX_SORT(6, 7); PIX_SORT(8, 9); PIX_SORT(10, 11);
PIX_SORT(12, 13); PIX_SORT(14, 15); PIX_SORT(1, 8); PIX_SORT(3, 10);
PIX_SORT(5, 12); PIX_SORT(7, 14); PIX_SORT(5, 8); PIX_SORT(7, 10);
return (p[7] + p[8]) / 2.;
}
static double opt_med25(double *p){
PIX_SORT(0, 1) ; PIX_SORT(3, 4) ; PIX_SORT(2, 4) ;
PIX_SORT(2, 3) ; PIX_SORT(6, 7) ; PIX_SORT(5, 7) ;
PIX_SORT(5, 6) ; PIX_SORT(9, 10) ; PIX_SORT(8, 10) ;
PIX_SORT(8, 9) ; PIX_SORT(12, 13); PIX_SORT(11, 13) ;
PIX_SORT(11, 12); PIX_SORT(15, 16); PIX_SORT(14, 16) ;
PIX_SORT(14, 15); PIX_SORT(18, 19); PIX_SORT(17, 19) ;
PIX_SORT(17, 18); PIX_SORT(21, 22); PIX_SORT(20, 22) ;
PIX_SORT(20, 21); PIX_SORT(23, 24); PIX_SORT(2, 5) ;
PIX_SORT(3, 6) ; PIX_SORT(0, 6) ; PIX_SORT(0, 3) ;
PIX_SORT(4, 7) ; PIX_SORT(1, 7) ; PIX_SORT(1, 4) ;
PIX_SORT(11, 14); PIX_SORT(8, 14) ; PIX_SORT(8, 11) ;
PIX_SORT(12, 15); PIX_SORT(9, 15) ; PIX_SORT(9, 12) ;
PIX_SORT(13, 16); PIX_SORT(10, 16); PIX_SORT(10, 13) ;
PIX_SORT(20, 23); PIX_SORT(17, 23); PIX_SORT(17, 20) ;
PIX_SORT(21, 24); PIX_SORT(18, 24); PIX_SORT(18, 21) ;
PIX_SORT(19, 22); PIX_SORT(8, 17) ; PIX_SORT(9, 18) ;
PIX_SORT(0, 18) ; PIX_SORT(0, 9) ; PIX_SORT(10, 19) ;
PIX_SORT(1, 19) ; PIX_SORT(1, 10) ; PIX_SORT(11, 20) ;
PIX_SORT(2, 20) ; PIX_SORT(2, 11) ; PIX_SORT(12, 21) ;
PIX_SORT(3, 21) ; PIX_SORT(3, 12) ; PIX_SORT(13, 22) ;
PIX_SORT(4, 22) ; PIX_SORT(4, 13) ; PIX_SORT(14, 23) ;
PIX_SORT(5, 23) ; PIX_SORT(5, 14) ; PIX_SORT(15, 24) ;
PIX_SORT(6, 24) ; PIX_SORT(6, 15) ; PIX_SORT(7, 16) ;
PIX_SORT(7, 19) ; PIX_SORT(13, 21); PIX_SORT(15, 23) ;
PIX_SORT(7, 13) ; PIX_SORT(7, 15) ; PIX_SORT(1, 9) ;
PIX_SORT(3, 11) ; PIX_SORT(5, 17) ; PIX_SORT(11, 17) ;
PIX_SORT(9, 17) ; PIX_SORT(4, 10) ; PIX_SORT(6, 12) ;
PIX_SORT(7, 14) ; PIX_SORT(4, 6) ; PIX_SORT(4, 7) ;
PIX_SORT(12, 14); PIX_SORT(10, 14); PIX_SORT(6, 7) ;
PIX_SORT(10, 12); PIX_SORT(6, 10) ; PIX_SORT(6, 17) ;
PIX_SORT(12, 17); PIX_SORT(7, 17) ; PIX_SORT(7, 10) ;
PIX_SORT(12, 18); PIX_SORT(7, 12) ; PIX_SORT(10, 18) ;
PIX_SORT(12, 20); PIX_SORT(10, 20); PIX_SORT(10, 12) ;
return (p[12]);
}
#undef PIX_SORT
#define PIX_SORT(a, b) {if (a > b) ELEM_SWAP(a, b);}
/**
* @brief quick_select - algorithm for approximate median calculation for array idata of size n
* @param idata (i) - input data array
* @param n - size of `idata`
* @return median value
*/
double quick_select(const double *idata, int n){
int low, high;
int median;
int middle, ll, hh;
double *arr = MALLOC(double, n);
memcpy(arr, idata, n*sizeof(double));
low = 0 ; high = n-1 ; median = (low + high) / 2;
for(;;){
if(high <= low) // One element only
break;
if(high == low + 1){ // Two elements only
PIX_SORT(arr[low], arr[high]) ;
break;
}
// Find median of low, middle and high doubles; swap into position low
middle = (low + high) / 2;
PIX_SORT(arr[middle], arr[high]) ;
PIX_SORT(arr[low], arr[high]) ;
PIX_SORT(arr[middle], arr[low]) ;
// Swap low double (now in position middle) into position (low+1)
ELEM_SWAP(arr[middle], arr[low+1]) ;
// Nibble from each end towards middle, swapping doubles when stuck
ll = low + 1;
hh = high;
for(;;){
do ll++; while (arr[low] > arr[ll]);
do hh--; while (arr[hh] > arr[low]);
if(hh < ll) break;
ELEM_SWAP(arr[ll], arr[hh]) ;
}
// Swap middle double (in position low) back into correct position
ELEM_SWAP(arr[low], arr[hh]) ;
// Re-set active partition
if (hh <= median) low = ll;
if (hh >= median) high = hh - 1;
}
double ret = arr[median];
FREE(arr);
return ret;
}
#undef PIX_SORT
#undef ELEM_SWAP
/**
* @brief calc_median - calculate median of array idata with size n
* the specific type of algorythm is choosen according to `n`
* @param idata (i) - input data array
* @param n - size of array `idata`
* @return median value
*/
double calc_median(const double *idata, int n){
if(!idata || n < 1){
WARNX(_("Wrong parameters"));
return 0.;
}
typedef double (*medfunc)(double *p);
medfunc fn = NULL;
const medfunc fnarr[] = {opt_med2, opt_med3, opt_med4, opt_med5, opt_med6,
opt_med7, opt_med8, opt_med9};
if(n == 1) return *idata;
if(n < 10) fn = fnarr[n - 2];
else if(n == 16) fn = opt_med16;
else if(n == 25) fn = opt_med25;
if(fn){
// copy data to new buffer - `idata` should leave unchanged
double *dataarr = MALLOC(double, n);
memcpy(dataarr, idata, sizeof(double)*n);
double medval = fn(dataarr);
FREE(dataarr);
return medval;
}else{
return quick_select(idata, n);
}
}
#define doubleLess(a,b) ((a)<(b))
#define doubleMean(a,b) (((a)+(b))/2)
typedef struct Mediator_t{
double* data; // circular queue of values
int* pos; // index into `heap` for each value
int* heap; // max/median/min heap holding indexes into `data`.
int N; // allocated size.
int idx; // position in circular queue
int ct; // count of doubles in queue
} Mediator;
/*--- Helper Functions ---*/
#define minCt(m) (((m)->ct-1)/2) //count of doubles in minheap
#define maxCt(m) (((m)->ct)/2) //count of doubles in maxheap
//returns 1 if heap[i] < heap[j]
static inline int mmless(Mediator* m, int i, int j){
return doubleLess(m->data[m->heap[i]],m->data[m->heap[j]]);
}
//swaps doubles i&j in heap, maintains indexes
static inline int mmexchange(Mediator* m, int i, int j){
int t = m->heap[i];
m->heap[i] = m->heap[j];
m->heap[j] = t;
m->pos[m->heap[i]] = i;
m->pos[m->heap[j]] = j;
return 1;
}
//swaps doubles i&j if i<j; returns true if swapped
static inline int mmCmpExch(Mediator* m, int i, int j){
return (mmless(m,i,j) && mmexchange(m,i,j));
}
//maintains minheap property for all doubles below i/2.
static void minSortDown(Mediator* m, int i){
for(; i <= minCt(m); i*=2){
if(i>1 && i < minCt(m) && mmless(m, i+1, i)) ++i;
if(!mmCmpExch(m,i,i/2)) break;
}
}
//maintains maxheap property for all doubles below i/2. (negative indexes)
static void maxSortDown(Mediator* m, int i){
for(; i >= -maxCt(m); i*=2){
if(i<-1 && i > -maxCt(m) && mmless(m, i, i-1)) --i;
if(!mmCmpExch(m,i/2,i)) break;
}
}
//maintains minheap property for all doubles above i, including median
//returns true if median changed
static int minSortUp(Mediator* m, int i){
while (i > 0 && mmCmpExch(m, i, i/2)) i /= 2;
return (i == 0);
}
//maintains maxheap property for all doubles above i, including median
//returns true if median changed
static int maxSortUp(Mediator* m, int i){
while (i < 0 && mmCmpExch(m, i/2, i)) i /= 2;
return (i == 0);
}
/*--- Public Interface ---*/
//creates new Mediator: to calculate `ndoubles` running median.
//mallocs single block of memory, caller must free.
static Mediator* MediatorNew(int ndoubles){
int size = sizeof(Mediator) + ndoubles*(sizeof(double)+sizeof(int)*2);
Mediator* m = malloc(size);
m->data = (double*)(m + 1);
m->pos = (int*) (m->data + ndoubles);
m->heap = m->pos + ndoubles + (ndoubles / 2); //points to middle of storage.
m->N = ndoubles;
m->ct = m->idx = 0;
while (ndoubles--){ //set up initial heap fill pattern: median,max,min,max,...
m->pos[ndoubles] = ((ndoubles+1)/2) * ((ndoubles&1)? -1 : 1);
m->heap[m->pos[ndoubles]] = ndoubles;
}
return m;
}
//Inserts double, maintains median in O(lg ndoubles)
static void MediatorInsert(Mediator* m, double v){
int isNew=(m->ct<m->N);
int p = m->pos[m->idx];
double old = m->data[m->idx];
m->data[m->idx]=v;
m->idx = (m->idx+1) % m->N;
m->ct+=isNew;
if(p>0){ //new double is in minHeap
if (!isNew && doubleLess(old,v)) minSortDown(m,p*2);
else if (minSortUp(m,p)) maxSortDown(m,-1);
}else if (p<0){ //new double is in maxheap
if (!isNew && doubleLess(v,old)) maxSortDown(m,p*2);
else if (maxSortUp(m,p)) minSortDown(m, 1);
}else{ //new double is at median
if (maxCt(m)) maxSortDown(m,-1);
if (minCt(m)) minSortDown(m, 1);
}
}
//returns median double (or average of 2 when double count is even)
static double MediatorMedian(Mediator* m){
double v = m->data[m->heap[0]];
if ((m->ct&1) == 0) v = doubleMean(v, m->data[m->heap[-1]]);
return v;
}
/*
// median + min/max
static double MediatorStat(Mediator* m, double *minval, double *maxval){
double v= m->data[m->heap[0]];
if ((m->ct&1) == 0) v = doubleMean(v,m->data[m->heap[-1]]);
double min = v, max = v;
int i;
for(i = -maxCt(m); i < 0; ++i){
int v = m->data[m->heap[i]];
if(v < min) min = v;
}
*minval = min;
for(i = 1; i <= minCt(m); ++i){
int v = m->data[m->heap[i]];
if(v > max) max = v;
}
*maxval = max;
return v;
}*/
// TODO: add adaptive filtering
/**
* @brief get_adp_median_cross - adaptive median filter by cross 3x3
* We have 5 datapoints and 4 inserts @ each step, so it's better to use opt_med5 instead of Mediator
* @param img (i) - input image
* @param out (o) - output image (allocated outside)
* @param adp - TRUE for adaptive filtering and FALSE for regular
*/
static void get_adp_median_cross(const doubleimage *img, doubleimage *out, _U_ bool adp){
size_t w = img->width, h = img->height;
double *med = out->data, *inputima = img->data, *iptr;
#ifdef EBUG
double t0 = dtime();
#endif
OMP_FOR()
for(size_t x = 1; x < w - 1; ++x){
double buffer[5];
size_t curpix = x + w, // index of current pixel image arrays
y, ymax = h - 1;
for(y = 1; y < ymax; ++y, curpix += w){
double md, *I = &inputima[curpix]; //, Ival = *I;
memcpy(buffer, I - 1, 3*sizeof(double));
buffer[3] = I[-w]; buffer[4] = I[w];
md = opt_med5(buffer);
/*
if(adp){
double s, l;
s = DBL_EPSILON + MIN(buffer[0], buffer[1]);
l = MAX(buffer[3], buffer[4]) - DBL_EPSILON;
if(s < md && md < l){
if(s < Ival && Ival < l) med[curpix] = Ival;
else med[curpix] = md;
}else{
med[curpix] = adp_med_5by5(img, x, y);
}
}else */
med[curpix] = md;
}
}
// process borders & corners (without adaptive)
double buf[5];
// left top
buf[0] = inputima[0]; buf[1] = inputima[0];
buf[2] = inputima[1]; buf[3] = inputima[w];
buf[4] = inputima[w + 1];
med[0] = opt_med5(buf);
// right top
iptr = &inputima[w - 1];
buf[0] = iptr[0]; buf[1] = iptr[0];
buf[2] = iptr[-1]; buf[3] = iptr[w - 1];
buf[4] = iptr[w];
med[w - 1] = opt_med5(buf);
// left bottom
iptr = &inputima[(h - 1) * w];
buf[0] = iptr[0]; buf[1] = iptr[0];
buf[2] = iptr[-w]; buf[3] = iptr[1 - w];
buf[4] = iptr[1];
med[(h - 1) * w] = opt_med5(buf);
// right bottom
iptr = &inputima[h * w - 1];
buf[0] = iptr[0]; buf[1] = iptr[0];
buf[2] = iptr[-w-1]; buf[3] = iptr[-w];
buf[4] = iptr[-1];
med[h * w - 1] = opt_med5(buf);
// process borders without corners
// top
OMP_FOR(shared(med))
for(size_t x = 1; x < w - 1; ++x){
double *iptr = &inputima[x];
buf[0] = buf[1] = *iptr;
buf[2] = iptr[-1]; buf[3] = iptr[2];
buf[4] = iptr[w];
med[x] = opt_med5(buf);
}
// bottom
size_t curidx = (h-2)*w;
OMP_FOR(shared(curidx, med))
for(size_t x = 1; x < w - 1; --x){
double *iptr = &inputima[curidx + x];
buf[0] = buf[1] = *iptr;
buf[2] = iptr[-w]; buf[3] = iptr[-1];
buf[4] = iptr[1];
med[curidx + x] = opt_med5(buf);
}
// left
OMP_FOR(shared(med))
for(size_t y = 1; y < h - 1; ++y){
size_t cur = y * w;
double *iptr = &inputima[cur];
buf[0] = buf[1] = *iptr;
buf[2] = iptr[-w]; buf[3] = iptr[1];
buf[4] = iptr[w];
med[cur] = opt_med5(buf);
}
// right
curidx = w - 1;
OMP_FOR(shared(curidx, med))
for(size_t y = 1; y < h - 1; ++y){
size_t cur = curidx + y * w;
double *iptr = &inputima[cur];
buf[0] = buf[1] = *iptr;
buf[2] = iptr[-w]; buf[3] = iptr[-1];
buf[4] = iptr[w];
med[cur] = opt_med5(buf);
}
DBG("time for median filtering by cross 3x3 of image %zdx%zd: %gs", w, h,
dtime() - t0);
}
// TODO: add borders and corners
/**
* @brief get_median - filter image by median (radius*2 + 1) x (radius*2 + 1)
* @param img (i) - input image
* @param radius - zone radius (0 for cross 3x3)
* @return image filtered by median (allocated here)
*/
doubleimage *get_median(const doubleimage *img, size_t radius){
size_t w = img->width, h = img->height;
doubleimage *out = doubleimage_new(img->width, img->height);
if(!out){
WARNX(_("Can't create output image"));
return NULL;
}
memcpy(out->data, img->data, sizeof(double)*img->totpix);
double *med = out->data, *inputima = img->data;
if(radius == 0){
get_adp_median_cross(img, out, 0);
return out;
}
size_t blksz = radius * 2 + 1, fullsz = blksz * blksz;
#ifdef EBUG
double t0 = dtime();
#endif
OMP_FOR(shared(inputima, med))
for(size_t x = radius; x < w - radius; ++x){
size_t xx, yy, xm = x + radius + 1, y, ymax = blksz - 1, xmin = x - radius;
Mediator* m = MediatorNew(fullsz);
// initial fill
for(yy = 0; yy < ymax; ++yy)
for(xx = xmin; xx < xm; ++xx)
MediatorInsert(m, inputima[xx + yy*w]);
ymax = 2*radius*w;
xmin += ymax;
xm += ymax;
ymax = h - radius;
size_t medidx = x + radius * w;
for(y = radius; y < ymax; ++y, xmin += w, xm += w, medidx += w){
for(xx = xmin; xx < xm; ++xx)
MediatorInsert(m, inputima[xx]);
med[medidx] = MediatorMedian(m);
}
FREE(m);
}
DBG("time for median filtering %zdx%zd of image %zdx%zd: %gs", blksz, blksz, w, h,
dtime() - t0);
return out;
}
#if 0
/**
* procedure for finding median value in window 5x5
* PROBLEM: bounds
*/
static double adp_med_5by5(const IMAGE *img, size_t x, size_t y){
size_t blocklen, w = img->width, h = img->height, yy, _2w = 2 * w;
double arr[25], *arrptr = arr, *dataptr, *currpix;
int position = ((x < 1) ? 1 : 0) // left columns
+ ((x > w - 2) ? 2 : 0) // right columns
+ ((y < 1) ? 4 : 0) // top rows
+ ((y > w - 2) ? 8 : 0); // bottom rows
/* Now by value of "position" we know where is the point:
***************************
* 5 * 4 * 6 *
***************************
* * * *
* * * *
* 1 * 0 * 2 *
* * * *
* * * *
***************************
* 9 * 8 *10 *
***************************/
currpix = &img->data[x + y * w]; // pointer to current pixel
dataptr = currpix - _2w - 2; // pointer to left upper corner of 5x5 square
inline void copy5times(double val){
for(int i = 0; i < 5; ++i) *arrptr++ = val;
}
inline void copy9times(double val){
for(int i = 0; i < 9; ++i) *arrptr++ = val;
}
void copycolumn(double *startpix){
for(int i = 0; i < 5; ++i, startpix += w) *arrptr++ = *startpix;
}
inline void copyvertblock(size_t len){
for(int i = 0; i < 5; ++i, dataptr += w, arrptr += len)
memcpy(arrptr, dataptr, len * sizeof(double));
}
inline void copyhorblock(size_t len){
for(size_t i = 0; i < len; ++i, dataptr += w, arrptr += 5)
memcpy(arrptr, dataptr, 5 * sizeof(double));
}
inline void copyblock(){
for(size_t i = 0; i < 4; ++i, dataptr += w, arrptr += 4)
memcpy(arrptr, dataptr, 4 * sizeof(double));
}
switch(position){
case 1: // left
copy5times(*currpix); // make 5 copies of current pixel
if(x == 0){ // copy 1st column too
dataptr += 2;
copycolumn(dataptr);
blocklen = 3;
}else{ // 2nd column - no copy need
++dataptr;
blocklen = 4;
}
copyvertblock(blocklen);
break;
case 2: // right
copy5times(*currpix);
if(x == w - 1){ // copy last column too
copycolumn(dataptr + 2);
blocklen = 3;
}else{ // 2nd column - no copy need
blocklen = 4;
}
copyvertblock(blocklen);
break;
case 4: // top
copy5times(*currpix);
if(y == 0){
dataptr += _2w;
memcpy(arrptr, dataptr, 5 * sizeof(double));
blocklen = 3;
}else{
dataptr += w;
blocklen = 4;
}
copyhorblock(blocklen);
break;
case 8: // bottom
copy5times(*currpix);
if(y == h - 1){
memcpy(arrptr, dataptr + _2w, 5 * sizeof(double));
blocklen = 3;
}else{
blocklen = 4;
}
copyhorblock(blocklen);
break;
case 5: // top left corner: in all corners we just copy 4x4 square & 9 times this pixel
copy9times(*currpix);
dataptr = img->data;
copyblock();
break;
case 6: // top right corner
copy9times(*currpix);
dataptr = &img->data[w - 4];
copyblock();
break;
case 9: // bottom left cornet
copy9times(*currpix);
dataptr = &img->data[(y - 4) * w];
copyblock();
break;
case 10: // bottom right cornet
copy9times(*currpix);
dataptr = &img->data[(y - 3) * w - 4];
copyblock();
break;
default: // 0
for(yy = 0; yy < 5; ++yy, dataptr += w, arrptr += 5)
memcpy(arrptr, dataptr, 5*sizeof(double));
}
return opt_med25(arr);
}
/**
* filter image by median (radius*2 + 1) x (radius*2 + 1)
*/
doubleimage *get_adaptive_median(const doubleimage *img, size_t radius);{
int radius = f->w;
size_t w = img->width, h = img->height, siz = w*h, bufsiz = siz*sizeof(double);
IMAGE *out = similarFITS(img, img->dtype);
double *med = out->data, *inputima = img->data;
memcpy(med, inputima, bufsiz);
if(radius == 0){
get_adp_median_cross(img, out, 1);
return out;
}
size_t blksz = radius * 2 + 1, fullsz = blksz * blksz;
#ifdef EBUG
double t0 = dtime();
#endif
OMP_FOR(shared(inputima, med))
for(size_t x = radius; x < w - radius; ++x){
size_t xx, yy, xm = x + radius + 1, y, ymax = blksz - 1, xmin = x - radius;
Mediator* m = MediatorNew(fullsz);
// initial fill
for(yy = 0; yy < ymax; ++yy)
for(xx = xmin; xx < xm; ++xx)
MediatorInsert(m, inputima[xx + yy*w]);
ymax = 2*radius*w;
xmin += ymax;
xm += ymax;
ymax = h - radius;
size_t curpos = x + radius * w;
for(y = radius; y < ymax; ++y, xmin += w, xm += w, curpos += w){
for(xx = xmin; xx < xm; ++xx)
MediatorInsert(m, inputima[xx]);
double s, l, md, I = inputima[curpos];
md = MediatorStat(m, &s, &l);
s += ITM_EPSILON, l -= ITM_EPSILON;
if(s < md && md < l){
if(s < I && I < l) med[curpos] = I;
else med[curpos] = md;
}else{
if(radius > LARGEST_ADPMED_RADIUS)
med[curpos] = I;
else
med[curpos] = adp_med_5by5(img, x, y);
}
}
FREE(m);
}
DBG("time for adadptive median filtering %zdx%zd of image %zdx%zd: %gs", blksz, blksz, w, h,
dtime() - t0);
return out;
}
#endif
Reference in New Issue View Git Blame Copy Permalink