Multi Scale Color Edge Detection Based on Vector Order Prewitt Operators
基于矢量Prewitt算子的多尺度彩色图象边缘检测方法

Many feature extraction methods begin with edge detection. Conventional color edge detection usually uses concepts similar to monochrome images. Consequently much color information are not thoroughly utilized. The idea of applying vector order statistics to edge detectors has been proposed to solve this problem. Because real world images contain distinct edges at various resolutions, effective extraction may require the combination of edges across several scales. Several researchers have proposed methods that employ filters at multiple scales. According to these approaches, we propose a scheme for multi scale color edge detection based on vector order Prewitt operators. Each channel of the color image is initially convolved with two Gaussian smoothing functions, one scale of that is higher and the other is lower. Then the vector order Prewitt operators are applied to the two smoothed images. Furthermore, we derive a rule based on the edge behavior in scale space to combine the two scale edge information. At last, the locations of the edge are detected after thinning and thresholding. Experiments have shown that the edges recovered by the proposed method are more integrated and accurate than those obtained by the vector order Prewitt operators running only at one scale. In addition, the edges detected by vector operators are much better than those detected using scale operators. It is indicated that using vector order statistics and multi scale approach is effective in color image edge detection.