Application of curvature driven diffusion model in lateral multi-lens video logging image inpainting

In order to restore the image´s defect which results from the lateral multi-lens video logging device. This paper proposes an image inpainting model algorithm based on curvature driven diffusion model which uses the thermal diffusion principle in Physics, diffusing the surrounding information to the region which needs to be repaired. Firstly, introduce the curvature factor to the diffusion coefficient of TV inpainting model. Secondly, discretize the divergence function. And according to the surrounding pixel information, calculate the defect pixel information which needs to be restored. Finally, fill the calculation results into the defect region of image to complete image defect restoring process. The experimental results show that the algorithm improves the connectivity deficiency of TV model, making the result image looks natural and satisfies the precision accuracy of image logging.