A New M Band and Watershed Combined Algorithm for Texture Segmentation of Multi Spectral Images
The M band wavelet decomposition, which is a direct generalisation of the standard 2 band wavelet decomposition, has been applied to the problem of segmentation of multi spectral images for its greater flexibility and better energy compaction which are required for good quality segmentation. The set of features derived from the wavelet domain are subjected to nonlinear filtering followed by smoothing to cause efficient representation of texture. Marker controlled watershed algorithm is applied on the gradient of the textural features to effectively segment the homogeneous regions. This new combined algorithm produces very good segmentation results for the textures in multi spectral images.
Keywords: M Band Wavelets, Local Energy Estimator, Watershed Algorithm, Multi Spectral Images
Assistant Professor, Electronics and Communication Engineering Department, Mepco Schlenk Engineering College, Sivakasi, Tamilnadu.
Dr. Ganesan Lakshmanan