New Directional Diffusion Algorithm Reconstructs Broken Images with Sharper Edges

A newly published study in Multimedia Tools and Applications introduces an algorithm designed to reconstruct missing or damaged regions of images by guiding diffusion along structural directions rather than spreading pixel values uniformly. The method, developed by Zohreh Rooygar, Ahmad Shirzadi, and Hossein Hosseinzadeh, targets a persistent weakness of classical inpainting approaches: the tendency to blur or distort curved boundaries where missing regions meet intact image content.

The core of the technique is an anisotropic diffusion equation adapted specifically for binary images. Instead of diffusing information equally in all directions, the algorithm performs a local gradient analysis around each point and steers the diffusion along the direction that follows existing edges. According to the authors, this edge-directed strategy allows curved boundaries to be reconstructed accurately while preserving the main structures of the image. A newly formulated diffusion coefficient gives the method fine control over how strongly information spreads, with the strength reduced in geometrically complex areas and near object boundaries where errors would be most damaging.

The binary version is only the first stage of the framework. The researchers extended it into a generalized, multi-stage pipeline for grayscale and color images. Importantly, the directional diffusion itself operates only in the binary stage. For grayscale and color reconstruction, the algorithm relies on the edges extracted earlier as barriers: these reconstructed contours block intensity from leaking across them, while isotropic diffusion fills in homogeneous regions on either side. The authors report that this barrier mechanism avoids the blurring and color artifacts that commonly appear near edges when diffusion-based inpainting is applied directly to color images.

The performance of the method was evaluated numerically on images containing complex, curved structures, and compared against several well-known inpainting techniques. The assessment used standard image quality metrics: Mean Squared Error (MSE), Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), Visual Information Fidelity (VIF), and Figure of Merit (FOM).

The reported results indicate that the proposed approach achieves high-accuracy restoration of damaged areas, preserves edges effectively, minimizes blurring in color images, and significantly reduces the number of computational iterations required.

The work was published in Multimedia Tools and Applications, volume 85, article number 827, with the article appearing on 29 October 2026. As with any algorithmic study, the reported advantages reflect the specific test conditions described by the authors, namely images with complex and curved structures evaluated with the listed quality metrics; broader performance across other image types would depend on further independent evaluation.

Subject of Research: An anisotropic diffusion-based image inpainting algorithm for reconstructing binary, grayscale, and color images.

Article Title: A new and efficient algorithm for directional diffusion-based reconstruction of binary images with extension to grayscale and color images

Article References: Rooygar, Z., Shirzadi, A., & Hosseinzadeh, H. (2026). A new and efficient algorithm for directional diffusion-based reconstruction of binary images with extension to grayscale and color images. Multimedia Tools and Applications, 85(11), Article 827. https://doi.org/10.1007/s11042-026-21978-0

Image Credits: AI Generated

DOI: 10.1007/s11042-026-21978-0

Keywords: image inpainting, anisotropic diffusion, binary images, grayscale images, color images, edge preservation, PSNR, SSIM, computer vision, image reconstruction, efficient, algorithm