#include <processor.h>
Inheritance diagram for EMAN::BilateralProcessor:
Public Member Functions | |
void | process_inplace (EMData *image) |
To process an image in-place. | |
string | get_name () const |
Get the processor's name. | |
string | get_desc () const |
Get the descrition of this specific processor. | |
TypeDict | get_param_types () const |
Get processor parameter information in a dictionary. | |
Static Public Member Functions | |
static Processor * | NEW () |
Static Public Attributes | |
static const string | NAME = "bilateral" |
Bilateral processing does non-linear weighted averaging processing within a certain window.
distance_sigma | means how large the voxel has impact on its neighbors in spatial domain. The larger it is, the more blurry the resulting image. | |
value_sigma | eans how large the voxel has impact on its in range domain. The larger it is, the more blurry the resulting image. | |
niter | how many times to apply this processing on your data. | |
half_width | processing window size = (2 * half_widthh + 1) ^ 3. |
Definition at line 3902 of file processor.h.
string EMAN::BilateralProcessor::get_desc | ( | ) | const [inline, virtual] |
Get the descrition of this specific processor.
This function must be overwritten by a subclass.
Implements EMAN::Processor.
Definition at line 3911 of file processor.h.
03912 { 03913 return "Bilateral processing on 2D or 3D volume data. Bilateral processing does non-linear weighted averaging processing within a certain window. "; 03914 }
string EMAN::BilateralProcessor::get_name | ( | ) | const [inline, virtual] |
Get the processor's name.
Each processor is identified by a unique name.
Implements EMAN::Processor.
Definition at line 3906 of file processor.h.
References NAME.
03907 { 03908 return NAME; 03909 }
TypeDict EMAN::BilateralProcessor::get_param_types | ( | ) | const [inline, virtual] |
Get processor parameter information in a dictionary.
Each parameter has one record in the dictionary. Each record contains its name, data-type, and description.
Reimplemented from EMAN::Processor.
Definition at line 3921 of file processor.h.
References EMAN::EMObject::FLOAT, EMAN::EMObject::INT, and EMAN::TypeDict::put().
03922 { 03923 TypeDict d; 03924 d.put("distance_sigma", EMObject::FLOAT, "means how large the voxel has impact on its neighbors in spatial domain. The larger it is, the more blurry the resulting image."); 03925 d.put("value_sigma", EMObject::FLOAT, "means how large the voxel has impact on its in range domain. The larger it is, the more blurry the resulting image."); 03926 d.put("niter", EMObject::INT, "how many times to apply this processing on your data."); 03927 d.put("half_width", EMObject::INT, "processing window size = (2 * half_widthh + 1) ^ 3."); 03928 return d; 03929 }
static Processor* EMAN::BilateralProcessor::NEW | ( | ) | [inline, static] |
void BilateralProcessor::process_inplace | ( | EMData * | image | ) | [virtual] |
To process an image in-place.
For those processors which can only be processed out-of-place, override this function to just print out some error message to remind user call the out-of-place version.
image | The image to be processed. |
Implements EMAN::Processor.
Definition at line 3869 of file processor.cpp.
References EMAN::EMData::get_attr(), EMAN::EMData::get_data(), EMAN::EMData::get_xsize(), EMAN::EMData::get_ysize(), EMAN::EMData::get_zsize(), LOGWARN, EMAN::Processor::params, square, and EMAN::EMData::update().
03870 { 03871 if (!image) { 03872 LOGWARN("NULL Image"); 03873 return; 03874 } 03875 03876 float distance_sigma = params["distance_sigma"]; 03877 float value_sigma = params["value_sigma"]; 03878 int max_iter = params["niter"]; 03879 int half_width = params["half_width"]; 03880 03881 if (half_width < distance_sigma) { 03882 LOGWARN("localwidth(=%d) should be larger than distance_sigma=(%f)\n", 03883 half_width, distance_sigma); 03884 } 03885 03886 distance_sigma *= distance_sigma; 03887 03888 float image_sigma = image->get_attr("sigma"); 03889 if (image_sigma > value_sigma) { 03890 LOGWARN("image sigma(=%f) should be smaller than value_sigma=(%f)\n", 03891 image_sigma, value_sigma); 03892 } 03893 value_sigma *= value_sigma; 03894 03895 int nx = image->get_xsize(); 03896 int ny = image->get_ysize(); 03897 int nz = image->get_zsize(); 03898 03899 if(nz==1) { //for 2D image 03900 int width=nx, height=ny; 03901 03902 int i,j,m,n; 03903 03904 float tempfloat1,tempfloat2,tempfloat3; 03905 int index1,index2,index; 03906 int Iter; 03907 int tempint1,tempint3; 03908 03909 tempint1=width; 03910 tempint3=width+2*half_width; 03911 03912 float* mask=(float*)calloc((2*half_width+1)*(2*half_width+1),sizeof(float)); 03913 float* OrgImg=(float*)calloc((2*half_width+width)*(2*half_width+height),sizeof(float)); 03914 float* NewImg=image->get_data(); 03915 03916 for(m=-(half_width);m<=half_width;m++) 03917 for(n=-(half_width);n<=half_width;n++) { 03918 index=(m+half_width)*(2*half_width+1)+(n+half_width); 03919 mask[index]=exp((float)(-(m*m+n*n)/distance_sigma/2.0)); 03920 } 03921 03922 //printf("entering bilateral filtering process \n"); 03923 03924 Iter=0; 03925 while(Iter<max_iter) { 03926 for(i=0;i<height;i++) 03927 for(j=0;j<width;j++) { 03928 index1=(i+half_width)*tempint3+(j+half_width); 03929 index2=i*tempint1+j; 03930 OrgImg[index1]=NewImg[index2]; 03931 } 03932 03933 // Mirror Padding 03934 for(i=0;i<height;i++){ 03935 for(j=0;j<half_width;j++) OrgImg[(i+half_width)*tempint3+(j)]=OrgImg[(i+half_width)*tempint3+(2*half_width-j)]; 03936 for(j=0;j<half_width;j++) OrgImg[(i+half_width)*tempint3+(j+width+half_width)]=OrgImg[(i+half_width)*tempint3+(width+half_width-j-2)]; 03937 } 03938 for(i=0;i<half_width;i++){ 03939 for(j=0;j<(width+2*half_width);j++) OrgImg[i*tempint3+j]=OrgImg[(2*half_width-i)*tempint3+j]; 03940 for(j=0;j<(width+2*half_width);j++) OrgImg[(i+height+half_width)*tempint3+j]=OrgImg[(height+half_width-2-i)*tempint3+j]; 03941 } 03942 03943 //printf("finish mirror padding process \n"); 03944 //now mirror padding have been done 03945 03946 for(i=0;i<height;i++){ 03947 //printf("now processing the %d th row \n",i); 03948 for(j=0;j<width;j++){ 03949 tempfloat1=0.0; tempfloat2=0.0; 03950 for(m=-(half_width);m<=half_width;m++) 03951 for(n=-(half_width);n<=half_width;n++){ 03952 index =(m+half_width)*(2*half_width+1)+(n+half_width); 03953 index1=(i+half_width)*tempint3+(j+half_width); 03954 index2=(i+half_width+m)*tempint3+(j+half_width+n); 03955 tempfloat3=(OrgImg[index1]-OrgImg[index2])*(OrgImg[index1]-OrgImg[index2]); 03956 03957 tempfloat3=mask[index]*(1.0f/(1+tempfloat3/value_sigma)); // Lorentz kernel 03958 //tempfloat3=mask[index]*exp(tempfloat3/Sigma2/(-2.0)); // Guassian kernel 03959 tempfloat1+=tempfloat3; 03960 03961 tempfloat2+=tempfloat3*OrgImg[(i+half_width+m)*tempint3+(j+half_width+n)]; 03962 } 03963 NewImg[i*width+j]=tempfloat2/tempfloat1; 03964 } 03965 } 03966 Iter++; 03967 } 03968 03969 //printf("have finished %d th iteration\n ",Iter); 03970 // doneData(); 03971 free(mask); 03972 free(OrgImg); 03973 // end of BilaFilter routine 03974 03975 } 03976 else { //3D case 03977 int width = nx; 03978 int height = ny; 03979 int slicenum = nz; 03980 03981 int slice_size = width * height; 03982 int new_width = width + 2 * half_width; 03983 int new_slice_size = (width + 2 * half_width) * (height + 2 * half_width); 03984 03985 int width1 = 2 * half_width + 1; 03986 int mask_size = width1 * width1; 03987 int old_img_size = (2 * half_width + width) * (2 * half_width + height); 03988 03989 int zstart = -half_width; 03990 int zend = -half_width; 03991 int is_3d = 0; 03992 if (nz > 1) { 03993 mask_size *= width1; 03994 old_img_size *= (2 * half_width + slicenum); 03995 zend = half_width; 03996 is_3d = 1; 03997 } 03998 03999 float *mask = (float *) calloc(mask_size, sizeof(float)); 04000 float *old_img = (float *) calloc(old_img_size, sizeof(float)); 04001 04002 float *new_img = image->get_data(); 04003 04004 for (int p = zstart; p <= zend; p++) { 04005 int cur_p = (p + half_width) * (2 * half_width + 1) * (2 * half_width + 1); 04006 04007 for (int m = -half_width; m <= half_width; m++) { 04008 int cur_m = (m + half_width) * (2 * half_width + 1) + half_width; 04009 04010 for (int n = -half_width; n <= half_width; n++) { 04011 int l = cur_p + cur_m + n; 04012 mask[l] = exp((float) (-(m * m + n * n + p * p * is_3d) / distance_sigma / 2.0f)); 04013 } 04014 } 04015 } 04016 04017 int iter = 0; 04018 while (iter < max_iter) { 04019 for (int k = 0; k < slicenum; k++) { 04020 size_t cur_k1 = (size_t)(k + half_width) * new_slice_size * is_3d; 04021 int cur_k2 = k * slice_size; 04022 04023 for (int i = 0; i < height; i++) { 04024 int cur_i1 = (i + half_width) * new_width; 04025 int cur_i2 = i * width; 04026 04027 for (int j = 0; j < width; j++) { 04028 size_t k1 = cur_k1 + cur_i1 + (j + half_width); 04029 int k2 = cur_k2 + cur_i2 + j; 04030 old_img[k1] = new_img[k2]; 04031 } 04032 } 04033 } 04034 04035 for (int k = 0; k < slicenum; k++) { 04036 size_t cur_k = (k + half_width) * new_slice_size * is_3d; 04037 04038 for (int i = 0; i < height; i++) { 04039 int cur_i = (i + half_width) * new_width; 04040 04041 for (int j = 0; j < half_width; j++) { 04042 size_t k1 = cur_k + cur_i + j; 04043 size_t k2 = cur_k + cur_i + (2 * half_width - j); 04044 old_img[k1] = old_img[k2]; 04045 } 04046 04047 for (int j = 0; j < half_width; j++) { 04048 size_t k1 = cur_k + cur_i + (width + half_width + j); 04049 size_t k2 = cur_k + cur_i + (width + half_width - j - 2); 04050 old_img[k1] = old_img[k2]; 04051 } 04052 } 04053 04054 04055 for (int i = 0; i < half_width; i++) { 04056 int i2 = i * new_width; 04057 int i3 = (2 * half_width - i) * new_width; 04058 for (int j = 0; j < (width + 2 * half_width); j++) { 04059 size_t k1 = cur_k + i2 + j; 04060 size_t k2 = cur_k + i3 + j; 04061 old_img[k1] = old_img[k2]; 04062 } 04063 04064 i2 = (height + half_width + i) * new_width; 04065 i3 = (height + half_width - 2 - i) * new_width; 04066 for (int j = 0; j < (width + 2 * half_width); j++) { 04067 size_t k1 = cur_k + i2 + j; 04068 size_t k2 = cur_k + i3 + j; 04069 old_img[k1] = old_img[k2]; 04070 } 04071 } 04072 } 04073 04074 size_t idx; 04075 for (int k = 0; k < slicenum; k++) { 04076 size_t cur_k = (k + half_width) * new_slice_size; 04077 04078 for (int i = 0; i < height; i++) { 04079 int cur_i = (i + half_width) * new_width; 04080 04081 for (int j = 0; j < width; j++) { 04082 float f1 = 0; 04083 float f2 = 0; 04084 size_t k1 = cur_k + cur_i + (j + half_width); 04085 04086 for (int p = zstart; p <= zend; p++) { 04087 size_t cur_p1 = (p + half_width) * (2 * half_width + 1) * (2 * half_width + 1); 04088 size_t cur_p2 = (k + half_width + p) * new_slice_size; 04089 04090 for (int m = -half_width; m <= half_width; m++) { 04091 size_t cur_m1 = (m + half_width) * (2 * half_width + 1); 04092 size_t cur_m2 = cur_p2 + cur_i + m * new_width + j + half_width; 04093 04094 for (int n = -half_width; n <= half_width; n++) { 04095 size_t k = cur_p1 + cur_m1 + (n + half_width); 04096 size_t k2 = cur_m2 + n; 04097 float f3 = Util::square(old_img[k1] - old_img[k2]); 04098 04099 f3 = mask[k] * (1.0f / (1 + f3 / value_sigma)); 04100 f1 += f3; 04101 size_t l1 = cur_m2 + n; 04102 f2 += f3 * old_img[l1]; 04103 } 04104 04105 idx = (size_t)k * height * width + i * width + j; 04106 new_img[idx] = f2 / f1; 04107 } 04108 } 04109 } 04110 } 04111 } 04112 iter++; 04113 } 04114 if( mask ) { 04115 free(mask); 04116 mask = 0; 04117 } 04118 04119 if( old_img ) { 04120 free(old_img); 04121 old_img = 0; 04122 } 04123 } 04124 04125 image->update(); 04126 }
const string BilateralProcessor::NAME = "bilateral" [static] |