Abstract
This paper analyzes evolutionary integral based methods for image restoration. They are multiscale linear models where the restored image evolves according to a Volterra equation, and the diffusion is handled by a convolution kernel. Well-posedness, scale-space properties, and long term behaviour are investigated for the continuous and semi-discrete models. Some numerical experiments are included. They provide different rules to select the kernel, and illustrate the performance of the evolutionary integral model in image denoising and contour detection.