Local Adaptive Bilateral Filter with Variation for Deblurring

Abstract

ABSTRACT In this study, alternative application of bilateral filter for image deblurring and enhancement is discovered. The concept of Total variation model is added to BF. Based on analyzing force distribution rules of variance, standard deviation is managed to distinguish degree of degrade. The optimization solution of total variation is gained by tracking minimum change channels and keep maximum edges. Experimentation proves that the new V-ABF can solve the deblurring problem where original BF is solution for de-noising. General Terms Pattern Recognition Keywords Variance adaptive bilateral filter, de-blurring, restoration, smoothing, sharpening, filtering, total variance 1. INTRODUCTION Image restoration, that is, the computation of the degree of blur and noise in a given image sequence and restore original image, is a well-known problem in image processing and has received significant attention in recent years. Total Variation (TV) is one example of a widely used approach to model image degrade. Numerous applications have been developed to by this model e.g., [8,9,10,11,12]. Solution of the model is optimization of keeping edges and limitation of change. While Bilateral Filter (BF) with duo Gaussian filters has been addressed in image de-noising as simple and effective filter, new developments of BF application were found e.g., [13,15,16]. This paper addresses to BF with reference of TV model for new development. Considering BF from global and local views, it is quite noticeable that relative change of Gaussian BF options creates dynamic range of vision effect. The longer the Gaussian standard deviations, the blur effect found more clearly. To improve the accuracy of the blur/edge estimation on an image suffered from degrading it would be helpful to use TV model. If the estimation of blur and noise is known, the noise and blur can be removed correctly. In fact, this estimation may be is covered under variation map, so variance can give control degree on restoration BF process for blur/noise image (cf. fig.2). The proposed framework is an adaptive bilateral filter, which includes explicit modeling of the restoration process as well image regularization terms, and is solved via efficient TV model. The input to model is an simple image for enhancement or degraded image with blur/noise for restoration, while the output are the corresponding enhanced/restored image, map of variance (fig.3c), local and global standard Gaussian deviation (fig.3b) and a comparison of BF and V-ABF output (fig.3d). As demonstrated in this paper, this join of estimation of variance, deviation regulation, and channel selection by minimization of change, are essential techniques when apply BF and TV concept. Before proceeding with the explicit description of the proposed framework, the color image in Fig. 1 is illustrated. It is a natural scene, very challenging and appropriate to demonstrate the advantage of the approach. In this figure, daisy become more flat after BF, but if get more clear edge by V-ABF filter.