Compressed Texton Based High Resolution Remote Sensing Image Classification

In order to avoid the high computational-complexity inherited in traditional texture extraction method, a new textural feature based on raw image pixel patch is proposed and it is applied to high resolution remote sensing image classification combining with support vector machine,First,the original texture extracted from local image patches are projected into the compressed sub-space using the random projection technique.Then,the texture dictionary which represents local features is learned with k-means in the compressed domain for each class.Then,the visual word map is formed by coding every texton in the samples to the nearest word in the texture dictionary,and then the histogram of the visual words map and the second moment of the words are fused as the global textural feature.At last,the global texture is put into the support vector machine to classification. The proposed method is proved to be effective for texture representation and improving accuracy by experiments on two images.