Source code for admit.util.filter.Filter2D

""" Filter2D --- 2-dimensional spectral filtering.
    ----------------------------------------------

    This module defines the 2D filter methods.
"""

import numpy as np
import math
try:
  import scipy.signal
except:
  print "WARNING: No scipy; Filter2D utility cannot function."

[docs]class Filter2D(object): """ This class defines and runs 2D spectral filters. The currently available filters are Gaussian, Hanning, Triangle, Welch, Boxcar, and Savitzky Golay. The output spectrum will be of the same length as the input spectrum, however some edge channels may be zeroed by some methods, depending on the input parameters. Parameters ---------- spec : numpy array 2D numpy array of the input spectrum (just the amplitudes). method : str The smoothing filter to apply: boxcar, gaussian, welch, hanning, triangle, or savgol. No default. Minimum matching is enabled with a minimum of 3 characters, i.e. box = boxcar. keyval : various Any keyword value pairs for the specific method chosen, see the notes for specific keywords. Attributes ---------- spec : numpy array The spectrum. len : int The length of the spectrum. methods : list A list of the available filters. [method]_args : dict A dictionary for each method giving its keywords and defaults (e.g. boxcar_args). method : str The method being used. Notes ----- Details of the different filter keywords and defaults: .. tabularcolumns:: |p{1.5cm}|p{2cm}|p{0.5cm}|p{8cm}| +------------+---------------+------+----------------------------------------------+ | Filter | Keyword | Def. | Description | +============+===============+======+==============================================+ | "boxcar" | "width" | 3 | Number of channels to average together | +------------+---------------+------+----------------------------------------------+ | "gaussian" | "width" | 7 | Number of channels to span with the gaussian | +------------+---------------+------+----------------------------------------------+ | "hanning" | "width" | 5 | Number of channels to include in the cos | +------------+---------------+------+----------------------------------------------+ | "triangle" | "width" | 5 | Number of channels to span with the triangle | +------------+---------------+------+----------------------------------------------+ | "welch" | "width" | 5 | Number of channels to use in the function | +------------+---------------+------+----------------------------------------------+ | "savgol" | "window_size" | 7 | Number of channels to use in the calculation | +------------+---------------+------+----------------------------------------------+ | | "order" | 3 | Order of the poynomial fit (must be odd) | +------------+---------------+------+----------------------------------------------+ | | "deriv" | 0 | The number of the derivative to compute | | | | | (0 = just smooth) | +------------+---------------+------+----------------------------------------------+ """ def __init__(self, data, method, **keyval): if len(data.shape) != 2: raise Exception("Spectrum is not 2D but you are trying to use a 2D filter.") self.data = data #self.len = self.spec.shape[0] self.methods = ["boxcar", "gaussian", "welch", "hanning", "triangle", "savgol"] # keywords for the different algorithms self.boxcar_args = {"width" : 5} self.gaussian_args = {"width" : 9} self.welch_args = {"width" : 7} self.hanning_args = {"width" : 7} self.triangle_args = {"width" : 7} self.savgol_args = {"window_size" : 11, "order" : 3, "deriv" : None} self.method = self.checkmethod(method) for k, v in keyval.iteritems(): try: a = getattr(self, method + "_args")[k] except: raise Exception("Unknown input %s for smoothing." % (k)) if type(a) != type(v): raise Exception("Cannot change the type of an attribute. %s must be a %s not a %s." % (k, type(a), type(v))) getattr(self, method + "_args")[k] = v
[docs] def checkmethod(self, method): """ Method to interpret the input method and determine the full method name Parameters ---------- method : str The method to use, minimal matching is possible, with a minimum of 3 characters (e.g. "box" will be interpreted to be "boxcar") Returns ------- None """ if len(method) < 3: raise Exception("Minimum of 3 characters are needed for minimal matching of strings.") for m in self.methods: if m.startswith(method): return m raise Exception("Unknown method %s given for smoothing. Available methods are: %s" % (method, str(self.methods)))
[docs] def radius(self, x , y, width): """ Method to calculate the radius of a point in the kernel Parameters ---------- x : float The x coordinate y : float The y coordinate width : int The width of the Gaussian being used Returns ------- Float containing the radius to the point """ return math.sqrt(math.pow(x - width / 2.0, 2) + math.pow(y - width / 2.0, 2))
[docs] def boxcar(self, width): r""" Method to apply a boxcar filter to a spectrum. The filter for point x[i] is defined as: .. math:: x[i] = \frac{1}{N} \sum_{n=0}^{N} x[i + n - \frac{N - 1}{2}] where N is the width of the filter. Parameters ---------- width : int The width of the box to use in channels, must be odd Returns ------- numpy array The smoothed image, (width - 1)/2 edge channels will be zeroed """ kernel = np.zeros((width, width)) for x in range(width): for y in range(width): kernel[y][x] = 1.0 kernel /= kernel.sum() return scipy.signal.convolve2d(self.data, kernel, boundary="symm")
[docs] def gaussian(self, width): r""" Method to apply a Gaussian filter to a spectrum. The filter for point x[i] is defined as: .. math:: x[i] = \frac{3}{N} \sum_{n=0}^{N} x[i + n - \frac{N - 1}{2}] e^{-\frac{1}{2}\left(\frac{n-(N-1)/2}{\sigma(N-1)/2}\right)^2} where N is the width of the filter. Parameters ---------- width : int The number of channels to span with the gaussian for each iteration, must be odd Returns ------- numpy array The smoothed image, (width - 1)/2 edge channels will be zeroed """ kernel = np.zeros((width, width)) for x in range(width): for y in range(width): kernel[y][x] = math.exp(-0.5 * pow(((self.radius(float(x), float(y), float(width)) - ((float(width) - 1.0) / 2.0)) / (0.2 * (float(width) - 1.0) / 2.0)), 2)) kernel /= kernel.sum() return scipy.signal.convolve2d(self.data, kernel, boundary="symm")
[docs] def welch(self, width): r""" Method to apply a Welch filter to a spectrum. The filter for point x[i] is defined as: .. math:: x[i] = \frac{3}{2(N-1)} \sum_{n=0}^{N} x[i + n - \frac{N - 1}{2}] \left(1 - \left(\frac{n - \frac{N-1}{2}}{\frac{N-1}{2}}\right)^2\right) where N is the width of the filter. Parameters ---------- width : int The number of channels to span with the function for each iteration, must be odd Returns ------- numpy array The smoothed image, (width - 1)/2 edge channels will be zeroed """ width += 2 # must add 2 to get the proper width kernel = np.zeros((width, width)) for x in range(width): for y in range(width): kernel[y][x] = (1 - pow((self.radius(float(y), float(x), float(width)) - (float(width - 1) / 2.0)) / (float(width - 1) / 2.0), 2)) kernel /= kernel.sum() return scipy.signal.convolve2d(self.data, kernel, boundary="symm")
[docs] def hanning(self, width): r""" Method to apply a Hanning filter to a spectrum. The filter for point x[i] is defined as: .. math:: x[i] = \frac{2}{N-1} \sum_{n=0}^{N} x[i + n - \frac{N - 1}{2}] 0.5 \left(1 - \cos\left(\frac{2\pi n}{N-1}\right)\right) where N is the width of the filter. Parameters ---------- width : int The number of channels to span with the function for each iteration, must be odd Returns ------- numpy array The smoothed image, (width - 1)/2 edge channels will be zeroed """ width += 2 # must add 2 to get the proper width kernel = np.zeros((width, width)) for x in range(width): for y in range(width): kernel[y][x] = 0.5 * (1.0 - math.cos((2.0 * math.pi * self.radius(float(y), float(x), float(width))) / float(width - 1))) kernel /= kernel.sum() return scipy.signal.convolve2d(self.data, kernel, boundary="symm")
[docs] def triangle(self, width): r""" Method to apply a Triangular filter to a spectrum. The filter for point x[i] is defined as: .. math:: x[i] = \sum_{n=0}^{N} x[i + n - \frac{N - 1}{2}] \left(1 - \left|\frac{n-\frac{N-1}{2}}{\frac{N}{2}}\right|\right) where N is the width of the filter. Parameters ---------- width : int The number of channels to span with the function for each iteration, must be odd Returns ------- numpy array The smoothed image, (width - 1)/2 edge channels will be zeroed """ kernel = np.zeros((width, width)) for x in range(width): for y in range(width): kernel[y][x] = (1.0 - abs((self.radius(float(y), float(x), float(width)) - (float(width - 1) / 2.0)) / (float(width) / 2.0))) kernel /= kernel.sum() return scipy.signal.convolve2d(self.data, kernel, boundary="symm")
[docs] def savgol(self, window_size, order, deriv=None): """ Method to apply a Savitzky-Golay filter to a 2D image. Parameters ---------- window_size : int the size of the window. Must be an odd integer number. order : int the order of the polynomial used in the filtering. Must be less then `window_size` - 1. deriv: int the order of the derivative to compute (default = None means only smoothing) Returns ------- numpy array The smoothed image, (width - 1)/2 edge channels will be zeroed """ # number of terms in the polynomial expression n_terms = (order + 1) * (order + 2) / 2.0 if window_size % 2 == 0: raise ValueError('window_size must be odd') if window_size ** 2 < n_terms: raise ValueError('order is too high for the window size') half_size = window_size // 2 # exponents of the polynomial. # p(x, y) = a0 + a1*x + a2*y + a3*x^2 + a4*y^2 + a5*x*y + ... # this line gives a list of two item tuple. Each tuple contains # the exponents of the k-th term. First element of tuple is for x # second element for y. # Ex. exps = [(0, 0), (1, 0), (0, 1), (2, 0), (1, 1), (0, 2), ...] exps = [(k - n, n) for k in range(order + 1) for n in range(k + 1)] # coordinates of points ind = np.arange(-half_size, half_size + 1, dtype=np.float64) dx = np.repeat(ind, window_size) dy = np.tile(ind, [window_size, 1]).reshape(window_size ** 2,) # build matrix of system of equation A = np.empty((window_size ** 2, len(exps))) for i, exp in enumerate(exps): A[:, i] = (dx ** exp[0]) * (dy ** exp[1]) # pad input array with appropriate values at the four borders new_shape = self.data.shape[0] + 2 * half_size, self.data.shape[1] + 2 * half_size Z = np.zeros((new_shape)) # top band band = self.data[0, :] Z[:half_size, half_size:-half_size] = \ band - np.abs(np.flipud(self.data[1:half_size + 1, :]) - band) # bottom band band = self.data[-1, :] Z[-half_size:, half_size:-half_size] = \ band + np.abs(np.flipud(self.data[-half_size - 1:-1, :]) - band) # left band band = np.tile(self.data[:, 0].reshape(-1, 1), [1, half_size]) Z[half_size:-half_size, :half_size] = \ band - np.abs(np.fliplr(self.data[:, 1:half_size + 1]) - band) # right band band = np.tile(self.data[:, -1].reshape(-1, 1), [1, half_size]) Z[half_size:-half_size, -half_size:] = \ band + np.abs(np.fliplr(self.data[:, -half_size - 1:-1]) - band) # central band Z[half_size:-half_size, half_size:-half_size] = self.data # top left corner band = self.data[0, 0] Z[:half_size, :half_size] = \ band - np.abs(np.flipud(np.fliplr(self.data[1:half_size + 1, 1:half_size + 1])) - band) # bottom right corner band = self.data[-1, -1] Z[-half_size:, -half_size:] = \ band + np.abs(np.flipud(np.fliplr(self.data[-half_size - 1:-1, -half_size - 1:-1])) - band) # top right corner band = Z[half_size, -half_size:] Z[:half_size, -half_size:] = \ band - np.abs(np.flipud(Z[half_size + 1:2 * half_size + 1, -half_size:]) - band) # bottom left corner band = Z[-half_size:, half_size].reshape(-1, 1) Z[-half_size:, :half_size] = \ band - np.abs(np.fliplr(Z[-half_size:, half_size + 1:2 * half_size + 1]) - band) # solve system and convolve if deriv is None: m = np.linalg.pinv(A)[0].reshape((window_size, -1)) return scipy.signal.fftconvolve(Z, m, mode='valid') elif deriv == 'col': c = np.linalg.pinv(A)[1].reshape((window_size, -1)) return scipy.signal.fftconvolve(Z, -c, mode='valid') elif deriv == 'row': r = np.linalg.pinv(A)[2].reshape((window_size, -1)) return scipy.signal.fftconvolve(Z, -r, mode='valid') elif deriv == 'both': c = np.linalg.pinv(A)[1].reshape((window_size, -1)) r = np.linalg.pinv(A)[2].reshape((window_size, -1)) return scipy.signal.fftconvolve(Z, -r, mode='valid'), scipy.signal.fftconvolve(Z, -c, mode='valid')
[docs] def run(self): """ Method to run the selected filter on the data Parameters ---------- None Returns ------- The smoothed image """ return getattr(self, self.method)(**getattr(self, self.method + "_args"))