Menu
I am getting weird errors when I try to convert a black and white PIL image to a numpy array. An example of the code I am working with is below.
if image.mode != '1':
image = image.convert('1') #convert to B&W
data = np.array(image) #Have also tried np.asarray(image)
n_lines = data.shape[0] #number of raster passes
line_range = range(data.shape[1])
for l in range(n_lines):
# process one horizontal line of the image
line = data[l]
for n in line_range:
if line[n] == 1:
write_line_to(xl, z+scale*n, speed) #conversion to other program code
elif line[n] == 0:
run_to(xl, z+scale*n) #conversion to other program code
I have tried this using both array and asarray for the conversion, and gotten different errors. If I use array, then the data I get out is nothing like what I put in. It looks like several very shrunken partial images side by side, with the remainder of the image space filled in in black. If I use asarray, then the entirety of python crashes during the raster step (on a random line). If I work with a greyscale image ('L'), then neither of these errors occurs for either array or asarray.
Does anyone know what I am doing wrong? Is there something odd about the way PIL encodes B&W images, or something special I need to pass numpy to make it convert properly?
555
I believe you've found a bug in PIL! (or possibly in numpy, but I'd wager it's on the PIL side of things...)
@c's answer above gives one workaround (use im.getdata()), though I'm not sure why numpy.asarry(image) is segfaulting for him... (Old version of PIL and/or numpy, maybe?) It works for me, but produces gibberish on 1-bit PIL images (and works for everything else, I use it frequently!).
Another workaround is to convert the BW image back to grayscale (mode 'L') before converting to a numpy array.
Converting the BW image back to grayscale before converting to a numpy array seems to be faster, if speed is important.
In [35]: %timeit np.array(im_bw.convert('L')).astype(np.uint8)
10000 loops, best of 3: 28 us per loop
In [36]: %timeit np.reshape(im_bw.getdata(), im_bw.size)
10000 loops, best of 3: 57.3 us per loop
On a seperate note, if you're modifying the array contents in-place, be sure to use numpy.array instead of numpy.asarray, as the latter will create an array from the PIL image instance without copying memory, thus returning a read-only array. Just mentioning this because I'm using asarray() below...
Here's a standalone example which confirms the bug...
import numpy as np
import Image
x = np.arange(256, dtype=np.uint8).reshape((16,16))
print 'Created array'
print x
im = Image.fromarray(x)
print 'Vales in grayscale PIL image using numpy.asarray <-- Works as expected'
print np.asarray(im)
print 'Converted to BW PIL image...'
im_bw = im.convert('1')
print 'Values in BW PIL image, using Image.getdata() <-- Works as expected'
print ' (Not a simple threshold due to dithering...)'
# Dividing by 255 to make the comparison easier
print np.reshape(im_bw.getdata(), (16, 16)) / 255
print 'Values in BW PIL image using numpy.asarray() <-- Unexpected!'
print ' (Same occurs when using numpy.array() to copy and convert.)'
print np.asarray(im_bw).astype(np.uint8)
print 'Workaround, convert back to type "L" before array conversion'
print np.array(im_bw.convert('L')).astype(np.uint8) / 255
Which outputs:
Created array
[[ 0 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]
[112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127]
[128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143]
[144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159]
[160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175]
[176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191]
[192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207]
[208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223]
[224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239]
[240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255]]
Vales in grayscale PIL image using numpy.asarray <-- Works as expected
[[ 0 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]
[112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127]
[128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143]
[144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159]
[160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175]
[176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191]
[192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207]
[208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223]
[224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239]
[240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255]]
Converted to BW PIL image...
Values in BW PIL image, using Image.getdata() <-- Works as expected
(Not a simple threshold due to dithering...)
[[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0]
[0 0 1 0 1 0 1 0 1 0 0 0 1 0 0 0]
[0 1 0 0 0 0 0 0 0 0 1 0 0 1 0 1]
[0 0 0 1 0 1 0 1 0 1 0 1 0 0 0 0]
[1 0 1 0 1 0 1 0 1 0 0 0 1 1 0 1]
[0 1 0 1 0 0 1 0 0 1 1 0 1 0 1 0]
[1 0 1 0 1 0 1 1 0 1 0 1 0 1 0 1]
[0 1 0 1 0 1 0 1 0 1 1 0 1 1 0 1]
[1 1 0 1 1 1 1 0 1 1 0 1 1 0 1 1]
[1 0 1 1 0 1 0 1 1 0 1 1 0 1 1 0]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 0 1 0 1 1 0 1 1 0 1 1 1 0 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]]
Values in BW PIL image using numpy.asarray() <-- Unexpected!
(Same occurs when using numpy.array() to copy and convert.)
[[0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[0 1 1 1 1 1 1 0 0 0 0 0 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]]
Workaround, convert back to type "L" before array conversion
[[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0]
[0 0 1 0 1 0 1 0 1 0 0 0 1 0 0 0]
[0 1 0 0 0 0 0 0 0 0 1 0 0 1 0 1]
[0 0 0 1 0 1 0 1 0 1 0 1 0 0 0 0]
[1 0 1 0 1 0 1 0 1 0 0 0 1 1 0 1]
[0 1 0 1 0 0 1 0 0 1 1 0 1 0 1 0]
[1 0 1 0 1 0 1 1 0 1 0 1 0 1 0 1]
[0 1 0 1 0 1 0 1 0 1 1 0 1 1 0 1]
[1 1 0 1 1 1 1 0 1 1 0 1 1 0 1 1]
[1 0 1 1 0 1 0 1 1 0 1 1 0 1 1 0]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 0 1 0 1 1 0 1 1 0 1 1 1 0 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]]
Not sure about this line:
data = numpy.array(image)
In fact, that gives me a segfault. But I just tried the following, and it works fine:
import numpy
import Image
im = Image.open("some_photo.jpg")
im = im.convert("1")
pixels = im.getdata() # returns 1D list of pixels
n = len(pixels)
data = numpy.reshape(pixels, im.size) # turn into 2D numpy array
for row in data:
# do your processing
pass
# Check that the numpy array's data is good
im2 = Image.new("1", im.size)
im2.putdata(numpy.reshape(data, [n, 1]))
im2.show()
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With
