Photometric invariant edge detection based on improved color tensor

Color information is important for image edge detection for its ability of preventing information loss due to iso-luminance and allowing exploiting the photometric information, such as shade, shading and specular compared with luminance information. In this paper, we focused on photometric invariant edge detection and proposed a new color edge detection model which integrating our proposed new form of color tensor with photometric quasi-invariant model; we also proposed a new measure for edge response using tensor nature. Experiments show that our method is effective and more discriminative comparing to several traditional methods.