Abstract: A weighted multiscale fundamental form (WMFF) is proposed to make the multiscale gradient information of multivalued image suitable for a single-valued image. Taking into account of the difference between image components, we assign different weights to different components, and the weights are adaptive and correlated to all gradients of image components. The larger the gradient magnitude, the higher the corresponding weight. WMFF has a variety of applications: image fusion, color image enhancement, anisotropic diffusion and color image restoration. Compared with multiscale fundamental form (MFF) based and other techniques, WMFF based techniques have better performance.