A practical image inpainting algorithm based on a predator-prey model

초록

Image inpainting aims to restore the structure and texture of damaged areas, and plays an important role in cultural heritage preservation, medical imaging and film production. However, it is still a challenge for existing methods to balance restoration effectiveness and computational cost. Inspired by ecological dynamics, this paper introduces a predator-prey model into image inpainting. Using population evolution characteristics, it provides a novel perspective on the information evolution between damaged and undamaged regions. In addition, this method also incorporates the reaction-diffusion mechanism, which can effectively smooth details. Numerical experiments show that this method can successfully remove scratches, noise, and text occlusion. Furthermore, it is applicable not only to grayscale images but also extends to color images. Compared to other methods, we significantly reduce the required time while maintaining high-quality restoration results. This research provides an efficient and novel solution for image inpainting, and effectively achieves a balance between restoration quality and computational cost.

키워드