-
Notifications
You must be signed in to change notification settings - Fork 3
/
resize_fft.py
executable file
·123 lines (106 loc) · 3.53 KB
/
resize_fft.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
#!/usr/bin/env python3
# encoding: utf-8
""" A tutorial for using IPP/MKL C API functions in Python """
__author__ = 'Jing Xu'
__email__ = '[email protected]'
import math
import numpy as np
import cv2
from ctypes import *
class IppiSize(Structure):
_fields_ = [("width", c_int),
("height", c_int)]
class IppiRect(Structure):
_fields_ = [("x", c_int),
("y", c_int),
("width", c_int),
("height", c_int)]
class IppiPoint(Structure):
_fields_ = [("x", c_int),
("y", c_int)]
class MKL_Complex16(Structure):
_fields_ = [("real", c_double),
("imag", c_double)]
# Load dynamic libraries
ipp = cdll.LoadLibrary("./intel64/libipp_rt.so")
mkl = cdll.LoadLibrary("/opt/intel/system_studio_2018/compilers_and_libraries_2018.2.199/linux/mkl/lib/intel64/libmkl_rt.so")
def resize(img_src, img_dst):
ssize = IppiSize(img_src.shape[1], img_src.shape[0])
srect = IppiRect(0, 0, img_src.shape[1], img_src.shape[0])
dsize = IppiSize(img_dst.shape[1], img_dst.shape[0])
drect = IppiRect(0, 0, img_dst.shape[1], img_dst.shape[0])
specSize = c_int(0)
initBufSize = c_int(0)
ipp.ippiResizeGetSize_8u(ssize, dsize, 2, 0, byref(specSize), byref(initBufSize))
pSpec = ipp.ippsMalloc_8u(specSize)
ipp.ippiResizeLinearInit_8u(ssize, dsize, pSpec)
bufSize = c_int(0)
ipp.ippiResizeGetBufferSize_8u(pSpec, dsize, 3, byref(bufSize))
pBuffer = ipp.ippsMalloc_8u(bufSize)
p = IppiPoint(0, 0)
ipp.ippiResizeLinear_8u_C1R(img_src.ctypes.data_as(POINTER(c_ubyte)), img_src.shape[1], img_dst.ctypes.data_as(POINTER(c_ubyte)), img_dst.shape[1], p, dsize, 1, 0, c_void_p(pSpec), c_void_p(pBuffer))
ipp.ippsFree(pSpec)
ipp.ippsFree(pBuffer)
def fft(img_data):
height = img_data.shape[0]
width = img_data.shape[1]
cdouble_imgsize = c_double * img_data.size
ccomplex_imgsize = MKL_Complex16 * img_data.size
x_real = cdouble_imgsize()
x_out = ccomplex_imgsize()
x_fft = np.zeros((height, width, 1), dtype=np.float64)
# Configure FFT handler
hand = c_void_p()
clong2 = c_long * 2
clong3 = c_long * 3
N = clong2()
N[0] = height
N[1] = width
mkl.DftiCreateDescriptor(byref(hand), 36, 33, 2, N)
mkl.DftiSetValue(hand, 11, 44)
mkl.DftiSetValue(hand, 10, 39)
rs = clong3()
rs[0] = 0
rs[1] = width
rs[2] = 1
cs = clong3()
cs[0] = 0
cs[1] = int(width/2+1)
cs[2] = 1
mkl.DftiSetValue(hand, 12, rs)
mkl.DftiSetValue(hand, 13, cs)
mkl.DftiSetValue(hand, 4, c_double(1.0/img_data.size))
mkl.DftiCommitDescriptor(hand)
# Load image data from 8U to double array
for i in range(0, height):
for j in range(0, width):
x_real[i*width+j] = c_double(math.pow(-1, i+j) * img_data[i][j])
# Perform FFT calculation
mkl.DftiComputeForward(hand, x_real, x_out)
# Expand compressed FFT results into full matrix
for j in range(0, width):
for i in range(0, height):
if j < width/2+1:
val = MKL_Complex16()
val.real = x_out[i*int(width/2+1)+j].real
val.imag = x_out[i*int(width/2+1)+j].imag
x_fft[i][j] = math.log(math.sqrt(val.real*val.real+val.imag*val.imag))
else:
if i == 0:
x_fft[0][j] = x_fft[0][width-j]
else:
x_fft[i][j] = x_fft[height-i][width-j]
# Normalize FFT results for visualization
fft_min = x_fft.min()
fft_max = x_fft.max()
x_fft = 255.0 * (x_fft - fft_min) / (fft_max - fft_min)
return x_fft.astype(np.uint8)
img = cv2.imread("testimg.jpg", cv2.IMREAD_GRAYSCALE)
img_resize = np.zeros((int(img.shape[0]/2), int(img.shape[1]/2), 1), dtype=np.uint8)
resize(img, img_resize)
img_fft = fft(img_resize)
cv2.imshow('img', img)
cv2.imshow('img_resize', img_resize)
cv2.imshow('img_fft', img_fft)
cv2.waitKey(0)
cv2.destroyAllWindows()