This paper introduces a new approach to color image processing that operates on multiscale edge descriptions of color component images. The multiscale edges are computed in the wavelet domain and recorded using the Wavelet Transform Modulus Maxima representation. The resulting description provides a means for assessing the strength and significance of individual edges, with deference to edges that coincide between color components. This leads to an effective means for suppressing noise in color images. A method of edge sharpening is also demonstrated. Results show the viability of this new form of processing.
展开▼
