## Monday, May 23, 2011

### Making a MFISH image

The first trial with numpy was not very successful. After code simplification, I get a result:
 RGB image combining five monospectral images
The code is:
```# -*- coding: utf-8 -*-
"""
Created on Mon May 23 14:07:40 2011

@author: jean-pat
"""

import numpy as np
import os
import pylab
#from scipy import linalg as la
#from scipy import ndimage as nd
import scipy as sp
user=os.path.expanduser("~")
workdir=os.path.join(user,"Applications","ImagesTest","MFISH")
blue="Aqua.tif"
green="Green.tif"
gold="Gold.tif"
red="Red.tif"
frd="FarRed.tif"
complete_path=os.path.join(workdir,blue)
#
complete_path=os.path.join(workdir,green)
#
complete_path=os.path.join(workdir,gold)
#
complete_path=os.path.join(workdir,red)
#
complete_path=os.path.join(workdir,frd)

R=0.8*i5+0.15*i4+0.05*i3+0.00*i2+0.00*i1
V=0.00*i5+0.05*i4+0.05*i3+0.80*i2+0.10*i1
B=0.00*i5+0.00*i4+0.05*i3+0.15*i2+0.80*i1
shape=i5.shape
rgb = np.zeros((shape[0],shape[1],3),dtype=np.uint8)
rgb[:,:,0]=np.copy(R)
rgb[:,:,1]=np.copy(V)
rgb[:,:,2]=np.copy(B)
pylab.subplot(111,frameon=False, xticks=[], yticks=[])
pylab.imshow(rgb)
pylab.show()```

```# -*- coding: utf-8 -*-
"""
Created on Mon May 23 14:07:40 2011

@author: jean-pat
"""

import numpy as np
import os
import pylab
#from scipy import linalg as la
#from scipy import ndimage as nd
import scipy as sp
user=os.path.expanduser("~")
workdir=os.path.join(user,"Applications","ImagesTest","MFISH")
blue="Aqua.tif"
green="Green.tif"
gold="Gold.tif"
red="Red.tif"
frd="FarRed.tif"
complete_path=os.path.join(workdir,blue)
#
complete_path=os.path.join(workdir,green)
#
complete_path=os.path.join(workdir,gold)
#
complete_path=os.path.join(workdir,red)
#
complete_path=os.path.join(workdir,frd)
######################
multi_spectr = np.dstack([i1,i2,i3,i4,i5])
conv_matrix = np.array([
[0.8,0.15,0.05,0,0],
[0,0.05,0.05,0.8,0.1],
[0,0,0.05,0.15,0.8]
])

shape = multi_spectr.shape
multi_spectr = multi_spectr.reshape(shape[0]*shape[1],shape[2])
rgb = np.dot(multi_spectr,conv_matrix.T).reshape(shape[:2]+(3,))
rgb8=np.uint8(rgb)
########################
#R=0.8*i5+0.15*i4+0.05*i3+0.00*i2+0.00*i1
#V=0.00*i5+0.05*i4+0.05*i3+0.80*i2+0.10*i1
#B=0.00*i5+0.00*i4+0.05*i3+0.15*i2+0.80*i1
#shape=i5.shape
#rgb = np.zeros((shape[0],shape[1],3),dtype=np.uint8)
#rgb[:,:,0]=np.copy(R)
#rgb[:,:,1]=np.copy(V)
#rgb[:,:,2]=np.copy(B)
pylab.subplot(111,frameon=False, xticks=[], yticks=[])
pylab.imshow(rgb8)
pylab.show()
```

That yields:

Anonymous said...

R=0.8*i5+0.15*i4+0.05*i3+0.00*i2+0.00*i1
V=0.00*i5+0.05*i4+0.05*i3+0.80*i2+0.10*i1
B=0.00*i5+0.00*i4+0.05*i3+0.15*i2+0.80*i1
how this covariance matrix came? is it a standard matrix?

JP Pommier said...

The issue is to make a three components RGB image with a five spectral components image. For example to make the red component of the RGB image, I take 80% of the far red image (i5), 15% of the red image (i4) and so on such that the sum of the weights of each "slice" of the five component image, is equal to 100%. That's how I choose the matrix coef. I don't know if there's a standart way to do that, may be some "best" condition(s). May be PCA could do something.
I'm not enough fluent in linear algebra to answer, you may read the chapter on Hotelling transform p148-157 in Digital Image Processing from Gonzalez & Wood (Adisson-Wesley).

JP Pommier said...
This comment has been removed by the author.
JP Pommier said...
This comment has been removed by the author.
Jean-Patrick Pommier said...

information from scikits-image group: