Chunsun Zhang
0; len « 5;
—(len- 4); 5<len<t (3)
1.0; lenzt
The same weight is applied for the computation of the following flanking region similarity scores.
We do not only compute a similarity score of line geometric properties, but also of photometric and chromatic
properties of line flanking regions. The photometric and chromatic line region attributes include the median and
standard deviation of the L band, and median and scatter matrix for the chromatic components, i.e. (a, b) data. Note that
when these properties are computed, the outliers in the flanking regions of each band are removed; for details see
(Henricsson, 1996). We first compare the medians of L, a, and b band in left and right flanking regions for a pair of
lines in two images. The similarity measurement is defined as the ratio of minimum median divided by the maximum.
For example, the left region median similarity measurement of the L band for a line pair is computed as:
min[ median (left image ), median (right image )]
Cu = (4)
max[ median (left image ), median (right image )]
This computation is also applied to the right regions of L band. Similarly, the median similarity measurements of a and
b bands are obtained. Then, we average these scores of L, a, b bands as one region similarity measurement, i.e. we
obtain the left and right region similarity measurement C,and C, for a line pair. In our matching algorithm, we do not
assume that the line pair has same contrast; instead we only request that at least one side of the line pair they
demonstrate similar brightness. Thus, if both C, and C, are less than a predefined value, then the two lines are treated as
different lines, and we stop computing similarity scores for them. Finally, the similarity scores for the median C, and
C, are multiplied with the weight, where the weight is defined as in (3), i.e.
Vi = WC; For left region
Vu = WC, For right region
c11 c12
The chromatic property of a region in the (a, b) data is represented by the scatter matrix c= [7 pn
| We further
describe this property as an ellipse. The shape of the ellipse is determined by its axes and orientation, derived from the
scatter matrix. Thereby, the similarity of a region chromatic property can be achieved through comparing the shapes of
the respective ellipses. We compute the orientation and roundness of ellipse as below:
Dir = lese AN Q = 4.DeHC)
C11? «c222' tr^(C)
(5)
The similarity measurement of a chromatic property is performed by comparing ellipse orientation and roundness in the
left and right regions for a line pair. The similarity score of ellipse orientation is computed with a form similar to (2).
For ellipse roundness, we use a form similar to (1).
All the scores are from 0 to 1, and the total similarity score is the average of all scores. The similarity score computation
starts from the longer lines, while the very short ones (< 5 pixels) are ignored.
3.3 Structural Matching with Probability Relaxation
After performing similarity measurement computation, we construct a matching pool and attach a similarity score to
each line pair. However, one still has problems to determine the best matches. The difficulty comes first how to decide
on a threshold and how to treat the case when a line is broken or occluded. In addition, matching using a very local
comparison of line attributes does not necessarily give results consistent in a local neighbourhood. For this reason,
structural matching receives more and more attention in computer vision and photogrammetry.
Structural matching establishes a correspondence from the primitives of one structural description to the primitives of a
second structural description (Haralick and Shapiro, 1993). Several methods for structural matching were developed in
the past (Vosselman, 1992). In this paper, the structural matching for line correspondence is realized through probability
relaxation.
1012 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.