Texture Based Image Classification

Texture classification is one of the four problem domains in the field of texture analysis. Here the approach for classification is based on the segmentation procedure that is used. The task of unsupervised texture segmentation has been the challenge of intensive research which attempts to discriminate between regions which have different textures. This paper addresses the textured image classification issue using the statistical based approach for extracting the spatial structures. Further level set method is used to analyze the feature space in order to extract homogeneous regions. This method is able to retrieve the texture regions and recover the shape from texture information in images. The approach for classification is based on moment features for segmentation. Based on the type of moments we are going to classify the images, these images are particularly from brodatz album. This work is poised on using the difference in mean and the standard duration of various features such as orientation and contrast, further using the level sets method for extracting the homogeneous regions. The statistical features are extracted using a window of suitable size around each pixel. Here we are trying to explore the possibility of using windows of different shapes and their benefits in segmenting the textures. As this method uses the difference in the first order statistical quantity, it yields the results in lesser time and using of level sets makes this method a robust one