Abstract: The hierarchical multiresolution wavelet analysis in conjunction with the contextual information of the image extracted from GMRF results in local features of the image. The model parameters are estimated by the LMS algorithm, and these parameters constitute a feature set of the image. The segmentation is then implemented via fuzzy partitioning of the feature set with the QFCM algorithm. The adaptiveness of the approach has been tested by simulations on a wide range of images. The experiment shows that this technique is especially effective for those images with low object surrounding contrast or images with low SNR.