Novel Weighted Hybrid Variational Model For Image Noise Reduction

Image denoising as a pre-processing scheme plays a significant role in digital image analysis.Despite of camera performance and quality,an image enhancement is always desirable to extend its convenient use in image applications.The need for efficient denoising method has increased with the gigantic production of digital images and videos that are often taken in poor conditions.Images are frequently noised due to capturing turbulences or transmission disturbances.Noise can be inherited due to defective instruments and deformation in data acquisition process.Digital images may incur various noises from different sources.Removing this unwanted noise from the images is one of the crucial tasks in image analysis.Hence,denoising came into account as a notable issue because of the necessity of eliminating noise before its utilization in any applications.In denoising problems,the crucial challenge is to eliminate the noise while preserving real information and avoiding unexpected modification in the images.The performance of traditional denoising approaches is not effective enough.Thus,it is still an enduring research toward better denoising results.Since edge preservation is a complicated issue during denoising process,designing an appropriate regularizer for a given fidelity is often a crucial matter in real-life applications.As a part of pre-processing,image denoising is also a demanding area of research since noise reduction and image detail preservation need a trade-off.For classical denoising models,the convex total variation(TV)or some nonconvex regularizers are used to achieve the trade-off.However,the existing denoising performance is still inadequate.To overcome the limitations,firstly,this study proposes a new variational method for image denoising,where a new regularizer is designed to protect more geometric-structural detail of images from over-smoothing,and eliminating much noise simultaneously.Then,we attempt to design a more robust smoothing-term in energy functional so that it can minimize the possibility of discontinuity and distortion of image edge details.In this work,we introduce a new hybrid denoising approach that inherits the advantages of both convex and nonconvex regularizers.A novel weight function is designed to improve denoising performance as well as to attain better generalization.The proposed method encompasses with a novel weighted hybrid regularizer in variational framework to ensure a better trade-off between the noise reduction and image edge preservation.A new algorithm based on Chambolle's method and iteratively reweighting method is also proposed to solve the model efficiently.The experimental results ensure that the proposed hybrid denoising method can perform better than the classical convex,nonconvex regularizer-based denoising approaches,and some other methods.