Non-Equal Spacing Division of HSV Components for Wood Image Retrieval

To make the HSV color model more suitable for wood image retrieval, one hundred images of wood were analyzed to educe their visual spatial distribution laws in HSV color space. It was found that hue, saturation and value components of wood were all centralized distribution. Based on the histogram threshold segmentation and color perception, the proposal on non-equal spacing division of hue, saturation and value was educed, which showed that the hue component could be divided into nine non-equal interval bins, the saturation component could be divided into four bins, and the value component could also be divided into four bins. Consequently, the dimension of feature vectors was effectively compressed, and the computational complexity of features matching was reduced.