Optimisation-Based Image Grid Smoothing for SST Images - Advanced Concepts for Intelligent Vision Systems - page 236

view, the content in the image looks preserved. The eciency of this type of
mesh smoothing has been proven on our dataset and its performance may be
compared favourably to the other filtering techniques for satellite images.
The grid smoothing process is applied on the smoothed image and the results
are depicted in Figure 2a and Figure 2b. It may be seen that the concentration
of points in the region presenting an edge (thermal front) is greater than in the
other regions. As θ increases, the deformation of the grid increases, the weight of
the initial coordinates being decreased in the cost function. With a large value
of θ , the repartition of the points is smoother along the edges while the details in
the shape are lost. On the other hand, a small value of θ leads to greater details
in the shape recovered, at the expense of a sparser repartition of the points. In
any case, continuous region with a large number of points may be observed and
may be interpreted as the boundaries of meso-scale sea structures.
6Conluon
A common framework for data smoothing and grid smoothing was presented in
the paper. A cost function was introduced for each case and that the solution of
the minimisation is unique. Using the conjugate gradient method, the comput-
ing time is reasonable and large image may be processed. An extensive study
of the convergence and computing time of the method may be found in [13].
Multiple applications of the framework are possible and will be investigated in
future research. Improved edge detection, image enhancement and compression
are among them. The reconstruction of the image will also be investigated and
compared to other interpolation schemes like in [10] and [11]. Finally, the en-
hanced images will be fed into a variational data assimilation scheme (4D-Var
for example) to test their ability to forecast the evolution of the sea surface
temperature.
References
1. Belkin, I.M., O'reilly, J.E.: An algorithm for oceanic front detection in chlorophyll
and SST satellite imagery. Journal of Marine Systems 78(3), 319-326 (2009)
2. Huot, E., Herlin, I., Korotaev, G.: Assimilation of SST satellite images for estima-
tion of ocean circulation velocity. In: Geoscience and Remote Sensing Symposium,
pp. II847-II850 (2008)
3. Cayula, J.-F., Cornillon, P.: Cloud detection from a sequence of SST images, Re-
mote Sens. Environ. 55, 80-88 (1996)
4. Hai, J., Xiaomei, Y., Jianming, G., Zhenyu, G.: Automatic eddy extraction from
SST imagery using artificial neural network. In: Proceedings of the International
Archives of the Photogrammetry, Remote Sensing and Spatial Information Science,
Beijing (2008)
5. Lim, Jae, S.: Two-Dimensional Signal and Image Processing, p. 548. Prentice Hall,
Englewood Cliffs (1990) equations 9.44 - 9.46