Full text: Proceedings of the Symposium "From Analytical to Digital" (Part 1)

  
1. 0 
the non-orthogonality rotation matrix expressed as | | 
$^ 1l, 
because for small angles, sin 8 = 8 and cos $ = 1, However, 
this is not the Case for feature boundary matching. The 
non-orthogonality may be large, for two reasons: 
1l. As features grow, they may do so along non-orthogonal 
axes. 
2. As features rotate, the low resolution of the camera may 
introduce several degrees of pseudo non-orthogonality. 
Therefore, a rigorous solution Was utilized. Linearized 
equations take the form : 
AV + BA = £f 
where A and B are the matrices of the partial derivatives 
with respect to measurements and parameters, 4 is the vector 
of the transformation parameters, and f is the discrepancy 
vector. 
In addition, a set of constraint equations was added to the 
System. With the Proper flag set, the Solution could be 
reduced to a five parameter (§ = 0), a four parameter (8 = 
0, 5, = Sy), or a three parameter transformation (§ = 0, s, 
= 8, = 1). The full System of equations is then: 
AV + BA = f 
CA = g 
where C is the matrix of the partial derivatives with 
respect to the constraint equations, 
The boundaries were matched by considering row and column 
pixel values as X and y point coordinates in the 
transformation. The Software made an initial estimate of a 
common point defining a local coordinate system origin in 
both features to be matched, so that the boundary pixels in 
each feature would have a common ordering. The boundary 
fitting was iterated in order to improve the estimate of the 
common point. 
4 TESTS WITH A SINGLE ROTATED FILM 
4.1 Data Collection 
In order to test the boundary detection and feature 
transformation algorithms, a single cranial X-ray was 
selected and imaged four times. It was rotated 
approximately 10, 15, and 25 degrees between successive 
images, and also translated from 20 to 40 mm in X and Y. No 
attempt was made to measure exactly the amount of rotation 
Or translation. The series of images was designed to 
simulate Sequential images of a growing person, except that 
the features did not change in shape, only in location and 
orientation. Also, the 'rotation angles were larger than 
- 3168 - 
  
  
 
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.