from ..utils import *
from .ssim import *
import numpy as np
import scipy.ndimage
def compute_msssim(frame1, frame2, method='product'):
extend_mode = 'constant'
avg_window = np.array(gen_gauss_window(5, 1.5))
K_1 = 0.01
K_2 = 0.03
level = 5
weight1 = np.array([0.0448, 0.2856, 0.3001, 0.2363, 0.1333])
weight2 = weight1.copy()
weight2 /= np.sum(weight2)
downsample_filter = np.ones(2, dtype=np.float32)/2.0
im1 = frame1.astype(np.float32)
im2 = frame2.astype(np.float32)
overall_mssim1 = []
overall_mssim2 = []
for i in range(level):
mssim_array, ssim_map_array, mcs_array, cs_map_array = ssim_full(im1, im2, K_1 = K_1, K_2 = K_2, avg_window = avg_window)
mssim_array = mssim_array[0]
ssim_map_array = ssim_map_array[0]
mcs_array = mcs_array[0]
cs_map_array = cs_map_array[0]
filtered_im1 = scipy.ndimage.correlate1d(im1, downsample_filter, 0)
filtered_im1 = scipy.ndimage.correlate1d(filtered_im1, downsample_filter, 1)
filtered_im1 = filtered_im1[1:, 1:]
filtered_im2 = scipy.ndimage.correlate1d(im2, downsample_filter, 0)
filtered_im2 = scipy.ndimage.correlate1d(filtered_im2, downsample_filter, 1)
filtered_im2 = filtered_im2[1:, 1:]
im1 = filtered_im1[::2, ::2]
im2 = filtered_im2[::2, ::2]
if i != level-1:
overall_mssim1.append(mcs_array**weight1[i])
overall_mssim2.append(mcs_array*weight2[i])
if method == "product":
overall_mssim = np.product(overall_mssim1) * mssim_array
else:
overall_mssim = np.sum(overall_mssim2) + mssim_array
return overall_mssim
[docs]def msssim(referenceVideoData, distortedVideoData, method='product'):
"""Computes Multiscale Structural Similarity (MS-SSIM) Index. [#f1]_
Both video inputs are compared frame-by-frame to obtain T
MS-SSIM measurements on the luminance channel.
Parameters
----------
referenceVideoData : ndarray
Reference video, ndarray of dimension (T, M, N, C), (T, M, N), (M, N, C), or (M, N),
where T is the number of frames, M is the height, N is width,
and C is number of channels. Here C is only allowed to be 1.
distortedVideoData : ndarray
Distorted video, ndarray of dimension (T, M, N, C), (T, M, N), (M, N, C), or (M, N),
where T is the number of frames, M is the height, N is width,
and C is number of channels. Here C is only allowed to be 1.
method : str
Whether to use "product" (default) or to use "sum" for combing multiple scales into the single score.
Returns
-------
msssim_array : ndarray
The MS-SSIM results, ndarray of dimension (T,), where T
is the number of frames
References
----------
.. [#f1] Z. Wang, E. P. Simoncelli and A. C. Bovik, "Multi-scale structural similarity for image quality assessment," IEEE Asilomar Conference Signals, Systems and Computers, Nov. 2003.
"""
referenceVideoData = vshape(referenceVideoData)
distortedVideoData = vshape(distortedVideoData)
assert(referenceVideoData.shape == distortedVideoData.shape)
T, M, N, C = referenceVideoData.shape
assert C == 1, "MS-SSIM called with videos containing %d channels. Please supply only the luminance channel" % (C,)
assert (M >= 176) & (N >= 176), "You supplied a resolution of %dx%d. MS-SSIM can only be used with videos large enough having multiple scales. Please use only with resolutions >= 176x176." % (M, N)
scores = np.zeros(T, dtype=np.float32)
for t in range(T):
referenceFrame = referenceVideoData[t, :, :, 0].astype(np.float32)
distortedFrame = distortedVideoData[t, :, :, 0].astype(np.float32)
scores[t] = compute_msssim(referenceFrame, distortedFrame, method=method)
return scores
