High Dynamic Range Image Compression Based on Diffusion Tensor Manipulation Gradient Domain

In this paper,we propose an improved high dynamic range image compression algorithm by manipulating the gradient field of the luminance image using diffusion tensor. Specifically, we calculate diffusion tensor for each pixel to construct a new attenuation function, and interpolate and merge with each scaling factor of attenuation function in order to reliably detect all of the significant intensity transitions at multiple scales, then exploit gradient approximation to reconstruct the enhanced image by solving the Poisson equation. The experimental results show that the proposed method is capable of compressing dynamic range while preserving the image of fine details and avoiding common artifacts, such as halos, thus make image have more natural color and better visual effects.