Title :
Contextual image segmentation based on AdaBoost and Markov random fields
Author_Institution :
Hiroshima Univ., Japan
Abstract :
AdaBoost, a machine learning algorithm, is employed for classification of land-cover categories of geostatistical data. We assume that the posterior probability is given by the odds ratio due to loss functions. Further, land-cover categories are assumed to follow Markov random fields (MRF). Then, we derive a classifier by combining two posteriors based on AdaBoost and MRF through the iterative conditional modes. Our procedure is applied to benchmark data sets provided by IEEE GRSS Data Fusion Committee and shows an excellent performance.
Keywords :
Markov processes; benchmark testing; geophysical signal processing; geophysical techniques; image segmentation; iterative methods; statistical analysis; AdaBoost; IEEE GRSS Data Fusion Committee; Markov random fields; benchmark data; contextual image segmentation; geostatistical data; iterative conditional modes; loss functions; machine learning algorithms; Image segmentation; Iterative methods; Machine learning; Machine learning algorithms; Markov random fields; Neural networks; Probability; Statistical analysis; Support vector machine classification; Support vector machines;
Conference_Titel :
Geoscience and Remote Sensing Symposium, 2003. IGARSS '03. Proceedings. 2003 IEEE International
Print_ISBN :
0-7803-7929-2
DOI :
10.1109/IGARSS.2003.1294836
