A semi-supervised support vector machine for texture segmentation

A novel semi-supervised support vector machine (S³VM) algorithm is developed for segmentation of texture images. The method classifies the uniform-texture regions from the regions of boundaries. The various-order statistics of the textures within a two-dimensional sliding window are measured. A K-means algorithm is used to label the two clusters approximately in the overall image. However, only those clusters which are very close to the class centres are labelled. Therefore, at this stage we have both the training and the working sets. A non-linear S³VM is then developed to exploit both sets to classify the regions. It is demonstrated that the algorithm is robust and the misclassification error is negligible. However, there may be a minor misplacement of the edges.