Morioh
Login
Molly Spencer

Molly Spencer

10 months ago

How to Create Single Image Random Dot Stereograms with TensorFlow

TensorFlow-SIRDS

Single Image Random Dot Stereogram for Tensorflow

SIRDS is a means to present 3D data in a 2D image. It allows for scientific data display of a waterfall type of plot with no hidden lines due to perspective. Data is encoded to show data in the Z direction as depth.

This project is based upon: https://www.cs.waikato.ac.nz/~ihw/papers/94-HWT-SI-IHW-SIRDS-paper.pdf

Sections

  • SIRDS usage demo for TensorFlow Op
TensorFlow single_image_random_dot_stereograms (SIRDS) DEMO

In [15]:

import tensorflow as tf
import matplotlib.pyplot as plt    #not required for SIRDS
import matplotlib.image as mpimg   #not required for SIRDS
from tensorflow.contrib.image import single_image_random_dot_stereograms
%matplotlib inline
import numpy as np

Outputs a single image random dot stereogram for export via encode_PNG/JPG OP.

Given the 2-D tensor 'depth_values' with encoded Z values, this operation will encode 3-D data into a 2-D image. The output of this Op is suitable for the encode_PNG/JPG ops. Be careful with image compression as this may corrupt the encode 3-D data witin the image.

This Op is based upon: 'https://www.cs.waikato.ac.nz/~ihw/papers/94-HWT-SI-IHW-SIRDS-paper.pdf'

Example use which outputs a SIRDS image as picture_out.png:

img=[[1,2,3,3,2,1],
     [1,2,3,4,5,2],
     [1,2,3,4,5,3],
     [1,2,3,4,5,4],
     [6,5,4,4,5,5]]

session = tf.InteractiveSession()

sirds = single_image_random_dot_stereograms(img,convergence_dots_size=8,number_colors=256,normalize=True)

out = sirds.eval()

png = tf.image.encode_png(out).eval()

with open('picture_out.png', 'wb') as f:
    f.write(png)

Args:

depth_values: Z values of data to encode into 'output_data_window' window, 
  lower values are further away {0.0 floor(far), 1.0 ceiling(near) after normalization}, must be 2-D tensor
hidden_surface_removal: Activate hidden surface removal
convergence_dots_size: Black dot size in pixels to help view converge image, drawn on bottom of image
dots_per_inch: Output device in dots/inch
eye_separation: Separation between eyes in inches
mu: Depth of field, Fraction of viewing distance (eg. 1/3 = .3333)
normalize: Normalize input data to [0.0, 1.0] 
normalize_max: Fix MAX value for Normalization - if < MIN, autoscale
normalize_min: Fix MIN value for Normalization - if > MAX, autoscale
border_level: Value of bord in depth 0.0 {far} to 1.0 {near}
number_colors: 2 (Black & White),256 (grayscale), and Numbers > 256 (Full Color) are all that are supported currently
output_image_shape: Output size of returned image in X,Y, Channels 1-grayscale, 3 color (1024, 768, 1),
  channels will be updated to 3 if 'number_colors' > 256
output_data_window: Size of "DATA" window, must be equal to or smaller than 'output_image_shape', will be centered
  and use 'convergence_dots_size' for best fit to avoid overlap if possible

Retuns:

image: A tensor of size 'output_image_shape' with the encloded 'depth_values'

In [16]:

session = tf.InteractiveSession()

In [17]:

img=[[1,2,3,3,2,1],
     [1,2,3,4,5,2],
     [1,2,3,4,5,3],
     [1,2,3,4,5,4],
     [6,5,4,4,5,5]]

img=np.array(img)

sirds = single_image_random_dot_stereograms(img,mu=.3333,convergence_dots_size=8,hidden_surface_removal=True,
                                            number_colors=2,normalize=True,border_level=0.00)

imgplot = plt.imshow(img,cmap=plt.cm.gray)

In [18]:

out = sirds.eval()

png = tf.image.encode_png(out,compression=-1).eval()
print "2 Colors output, 1 color channel needed",out.shape

with open('picture_out_2Color.png', 'wb') as f:
    f.write(png)

2 Colors output, 1 color channel needed (768, 1024, 1)

2 Color SIRDS

In [19]:

sirds = single_image_random_dot_stereograms(img,mu=.3333,convergence_dots_size=8,hidden_surface_removal=True,
                                            number_colors=256,normalize=True,border_level=0.00)
out = sirds.eval()

png = tf.image.encode_png(out,compression=-1).eval()
print "256 Colors output, 1 color channel needed",out.shape

with open('picture_out_256Color.png', 'wb') as f:
    f.write(png)

256 Colors output, 1 color channel needed (768, 1024, 1)

256 Color SIRDS

In [20]:

sirds = single_image_random_dot_stereograms(img,mu=.3333,convergence_dots_size=8,hidden_surface_removal=True,
                                            output_image_shape=[1024,768,3],normalize=True,border_level=0.00)
#Same as the following for true color, number_colors>256
#sirds = single_image_random_dot_stereograms(img,mu=.3333,convergence_dots_size=8,hidden_surface_removal=True,
#                                            number_colors=257,normalize=True,border_level=0.00)
out = sirds.eval()

png = tf.image.encode_png(out,compression=-1).eval()
print "True Colors output, 3 color channels needed",out.shape

with open('picture_out_TrueColor.png', 'wb') as f:
    f.write(png)

True Colors output, 3 color channels needed (768, 1024, 3)

True Color SIRDS

In [21]:

sirds = single_image_random_dot_stereograms(img,mu=.3333,convergence_dots_size=20,hidden_surface_removal=True,
                                            number_colors=257,normalize=True,border_level=0.00,
                                            output_data_window=[640,480])
out = sirds.eval()

png = tf.image.encode_png(out,compression=-1).eval()
print "True Colors output, 3 color channels needed",out.shape

with open('picture_out_TrueColor_window.png', 'wb') as f:
    f.write(png)

True Colors output, 3 color channels needed (768, 1024, 3)

True Color SIRDS, 640x480 window with 1024x768 image, allows for large convergence dot size as well, set to 20

In [22]:

sirds = single_image_random_dot_stereograms(img,mu=.3333,convergence_dots_size=20,hidden_surface_removal=True,
                                            number_colors=257,normalize=True,border_level=0.50,
                                            output_data_window=[640,480])
out = sirds.eval()

png = tf.image.encode_png(out,compression=-1).eval()
print "True Colors output, 3 color channels needed",out.shape

with open('picture_out_TrueColor_window_50.png', 'wb') as f:
    f.write(png)

True Colors output, 3 color channels needed (768, 1024, 3)

True Color SIRDS, 640x480 window with 1024x768 image, allows for large convergence dot size as well, set to 20

Border Level has been raised to 50% of depth range, note that convergence dots might be more difficult to converge since they don't appear on the far plane or 0%

CONE like shape

In [23]:

# Create a CONE type shape for demo
img=range(0,500)+range(499,0,-1)

x = np.linspace(0,10,999)
img2=np.array(img)
coneImage=np.array(img)
coneImage=coneImage.reshape(1,-1)
coneImage=coneImage*coneImage.T

plotout=plt.plot(x,img2,'k--')

In [24]:

plotout=plt.imshow(coneImage,cmap=plt.cm.gray)

In [25]:

sirds = single_image_random_dot_stereograms(coneImage,mu=.3333,convergence_dots_size=20,hidden_surface_removal=True,
                                            number_colors=257,normalize=True,border_level=0.00,
                                            output_data_window=[640,480])

out = sirds.eval()

png = tf.image.encode_png(out,compression=-1).eval()
print "True Colors output, 3 color channels needed",out.shape

with open('picture_out_TrueColor_cone.png', 'wb') as f:
    f.write(png)

True Colors output, 3 color channels needed (768, 1024, 3)

Cone like image

In [26]:

sirds = single_image_random_dot_stereograms(coneImage,mu=.6666,convergence_dots_size=20,hidden_surface_removal=True,
                                            number_colors=257,normalize=True,border_level=0.00,
                                            output_data_window=[640,480])
out = sirds.eval()

png = tf.image.encode_png(out,compression=-1).eval()
print "True Colors output, 3 color channels needed",out.shape

with open('picture_out_TrueColor_cone_66.png', 'wb') as f:
    f.write(png)

True Colors output, 3 color channels needed (768, 1024, 3)

Cone like image with larger field of depth 66%

In [27]:

sirds = single_image_random_dot_stereograms(coneImage,mu=.6666,convergence_dots_size=20,hidden_surface_removal=True,
                                            number_colors=257,normalize=True,border_level=0.500,
                                            output_data_window=[640,480])
out = sirds.eval()

png = tf.image.encode_png(out,compression=-1).eval()
print "True Colors output, 3 color channels needed",out.shape
with open('picture_out_TrueColor_cone_66_border_50.png', 'wb') as f:
    f.write(png)

True Colors output, 3 color channels needed (768, 1024, 3)

Cone like image full color Border level 50%

In [28]:

sirds = single_image_random_dot_stereograms(coneImage,mu=.6666,convergence_dots_size=20,hidden_surface_removal=False,
                                            number_colors=257,normalize=True,border_level=1.00,
                                            output_data_window=[640,480])
out = sirds.eval()

png = tf.image.encode_png(out,compression=-1).eval()
print "True Colors output, 3 color channels needed",out.shape

with open('picture_out_TrueColor_cone_66_border_artifact.png', 'wb') as f:
    f.write(png)

True Colors output, 3 color channels needed (768, 1024, 3)

Cone like image with border artifact due to hidden surface FALSE

Download details:

Author: Mazecreator
Source: https://github.com/Mazecreator/TensorFlow-SIRDS

License: Apache-2.0 license

#tensorflow 

How to Create Single Image Random Dot Stereograms with TensorFlow
1.00 GEEK