Pixel classification is one basic issues of image processing. A novel method of pixel classification based on orientation information measure of image is presented. Firstly

the general characterisitics of the pixels belonging to various pixel calsses are discussed

and the radical principles of pixel classification are also given in this paper. Then

each pixel of a gray value image is classified into the edge

the smooth

or the textured feature point

based on the orientation information measure of image

and variance of the orientation information measure when the scale of observation is changed. The method effectively integrates the multi scaled observations of image and the information measure of image. Finally

the effectiveness of this method has been testified by some experiments in the end of this paper. The method can generally be used as a front end of image processing and image analyses

has the better capability for denoise.