Automatic segmentation of linear structures on AVHRR images
- Centre Mulitaire d`Oceanographie, Brest (France)
- CISI, Plouzane (France)
This paper presents a new technique for the segmentation of infrared sea surface images applied to detection of linear oceanic structures (sea surface thermal fronts). The method is based upon the Mumford and Shah criterion, which consists of the minimization of an energy function using multiscale considerations. This method allows the segmentation of the image in several homogeneous regions. The main interest of the method is that it is much more adaptive than a classical gradient technique, and that detected linear structures appears in a more continuous way. We used this method for the analysis of several images of the Bay of Biscay. Detected linear structures are automatically extracted, with different informations about the structure (position, mean sea temperature gradient,...). The results on some typical images are discussed.
- OSTI ID:
- 508179
- Report Number(s):
- CONF-960384--
- Country of Publication:
- United States
- Language:
- English
Similar Records
Automatic segmentation of liver structure in CT images
Variational and Shape Prior-based Level Set Model for Image Segmentation