In this paper a novel large target region image inpainting model which can preserve linear structure and texture information is proposed. The model uses a cross isophotes diffusion equation constrain the inpainting order. The cross isophotes equation diffuses considering the edge extent

so the model has good linear structure preserving property. Diffusion of cross isophotes is morphologically invariant and it can fill the target region of the natural scene very well. The euclidean distance is used as the similarity function

while the partial differential equation constraint which is used to enlarge the effect of pixels in linear structure is employed. To reduce the seams caused by this model

and achieve a seamless image

a total variation interpolation constrained by gradient is used in the inpainted result. At last the similarity function is extended

and the target region can be inpainted to the assigned texture. Both theoretical analysis and experiments have verified the validity of the new model.