Image Retrieval based on the Multiwavelets Texture-Spatial Features
Zhiyong An, Jinjiang Li, Zhao feng, Jie Guo · Applied Mathematics & Information Sciences · 2013
Abstract: A new retrieval algorithm based on the texture-spatial texture using the GHM multiwavelets is presented. In order to describe the important visual information of image, we design the improved multiwavelets quantization map that can depict the important visual information for the multiwavelets sub-bands. Furthermore, the visual spatial histogram of multiwavelets quantization map is used as the texture-spatial features that denote the global texture information of image. At the same time, the local binary pattern is used to describe the local texture-spatial feature for the low frequency multiwavelets sub-band. Finally, the similarity of visual spatial histogram is computed by the spatial-weighted distance. Experiments indicate that this method gives better performance than looseness texturespatial algorithm and MTH algorithm in the natural image retrieval.