Color image enhancement and denoising using an optimized filternet based Local Structure Tensor analysis

Tensor based orientation adaptive filtering constitutes a flexible framework for image enhancement. In this paper, Local Structure Tensor (LST) based Adaptive Anisotropic Filtering (AAF) methodology is used for color image enhancement and denoising. This filtering framework enhances and preserves important, typically anisotropic, image structures while suppressing high-frequency noise. The goal of this work is to reduce the overall computational cost with minimum risk on accuracy by introducing optimized filternets for local structure analysis and reconstruction. This filtering technique facilitates user interaction and direct control over high frequency contents of the signal. The efficacy of the filtering framework is evaluated by testing the system with Lena image along with Gaussian and speckle noise added images. The results are compared using three different quality measures. Experimental results show that a good level of noise reduction along with structure enhancement can be achieved in the adaptively filtered images.