matrice.py

Go to the documentation of this file.

from __future__ import division
from object import Surface as fc
import numpy as np
import math
 
 
def create_matrix_l(faces, sigma, val):
        # create matrix L as in Mura paper
        riga = []
        # create matrix with only first line
        face = faces[0]
        for face2 in faces:
                adj = fc.adjacent(face, face2)
                if face == face2:
                        val = 1
                else:
                        if adj:
                                e = fc.common_edge(face, face2)
                                w = e[0].weight
                                val = math.exp(-w/sigma)
                        else:
                                val = 0
                riga.append(val)
        matrix_l = np.matrix([riga])
        del riga[:]
 
        for face in faces[1:]:
                for face2 in faces:
                        adj = fc.adjacent(face, face2)
                        if face == face2:
                                val = 1
                        else:
                                if adj:
                                        e = fc.common_edge(face, face2)
                                        w = e[0].weight
                                        val = math.exp(-w/sigma)
                                else:
                                        val = 0
                        riga.append(val)
                matrix_l = np.vstack([matrix_l, riga])
                del riga[:]
        return matrix_l
 
 
def create_matrix_d(matrix_l):
        # create matrix D as in Mura paper
        vett = []
        for row in matrix_l:
                vett.append(row.sum())
        matrix_d = np.asmatrix(np.diag(vett))
        return matrix_d
 
 
def symmetry(M):
        # make symmetric matrix M to put in input at dbscan
        if type(M) == np.matrix:
                M = np.array(M)
        [r, c] = M.shape
        M2 = np.copy(M)
        for i in range(r):
                for j in range(i, c):
                        M2[i, j] = (M[i, j] + M[j, i]) / 2
                        M2[j, i] = M2[i, j]
        return M2