A modified texture image segmentation algorithm is proposed. Based on the assumption that the observed texture image is the sum of the segmentation image and the irregular corruptive noise, the texture image is modeled by a contaminated Gaussian distribution. By a multiresolution model, the corresponding parameters at lower resolution are computed. Then the multiresolution segmentation algorithm first segments texture images at coarse resolution and then progresses to finer resolutions until individual pixels are classified. In addition the use of neighborhood priori information is more reasonable. The method results in accurate segmentations and requires significantly less computation.