The Investigation Of The Image Inpainting Techniques

With the development of science and technology, and the improvement of people's living standard, digital image processing technology has been applied to many fields. It makes the technology of digital image processing become a critical field of scientific research. Image inpainting technique is an important aspect in the technology of digital image processing research, the main idea is to enable the observers not to find that the images have been modified, or to obtain some special effects, such as image objects hidden or text removal.This thesis introduces the concept of digital image processing, its application ranges, research signification and research status. The typical algorithm of digital image inpainting is divided into two parts to be introduced: The first part is based on the partial differential equation (PDE). Such algorithm can acquire better effect when applied to deal with much smaller damaged images. Its main model includes the BSCB model, the TV model, the CDD model and the Mumford-Shah model, and the other is on the basis of block texture synthesis, it acquires much better effect when used for big damaged images. This thesis also introduces the image quality evaluation criteria both from subjective and objective stands, and points out the evaluation criteria, used in this thesis.Also, this thesis introduces the classical total variation (TV) model in detail, and analyzes both the advantages and disadvantages of this algorithm. To improve the TV model, a weight function taking the gradient at the reference point as its independent variable is added. This makes the references points with different positions have different values with respect to the point to be fixed. It makes different smoothness in different areas so as to make the edges sharper, and transition more natural. At last, the analysis of the experimental data reflects the effectiveness of the improved algorithm.