Abstract: Aiming at images with a lot of missing information and scratches, this paper puts forward a
new restoration algorithm by combining the advantages of CDD model and Criminisi algorithm.
Specifically, this algorithm first adopts the equalized operation of detail-oriented and self-adaptive
histogram to expand the dynamic range of the images, and then utilizes morphology operators to
separate the scratch regions and large areas in the images that need to be restored. Subsequently, CDD
model is applied to restore the scratches and then the improved priority Criminisi algorithm is used to
implement subsequent restoration. Finally, the restored images corresponding to the pixel of missing
areas are filled into the regions of images need restoration. As the experimental results indicate, the
improved algorithm has overcome the possible deviation continuation problems in current algorithm
and can make the restored images more consistent with people’s visual effect.

Key words: CDD model, Criminisi algorithm, image inpainting, equalization of self-adaptive
histogram, morphology operators