This paper presents a novel framework for seamless image processing in gradient domain. A gradient field of image region to be processed is introduced into general total variation(TV) model, which gives an extended version of TV model called gradient field total variation (GFTV)model. Seamless processing is performed by minimizing the model function. Through our framework based on GFTV model, a variety of applications for seamless image processing are introduced, mainly including seamless importation of source image regions into destination image, modifying the appearance, e.g. texture, illumination, of the image seamlessly within a selected region, and image seamless stitching.