Study On Methods Of Color Image Filtering

Images are often corrupted by noise during the acquisition or transmission process. So noise cancellation/filtering is an important task in image processing, especially when the final product is used for edge detection, image segmentation, and data compression. In recent years, significant advances have been made in the development of noise attenuation techniques for multichannel images.Image signals are composed of flat regional parts and abrupt changing areas, such as edges, which carry important information in visual perception. So it is important to preserve edges or details while attenuating noise in images. In this thesis, we first introduce the conventional methods of color image filtering. Then, we propose some new algorithms to reduce noise efficiently and at the same, to preserve original pixel values.(1) As to the color images corrupted by impulse noise, it is obvious that not all pixels are corrupted by impulse noise. Although the conventional filters may possess good impulsive noise attenuation characteristics, their performance is often accompanied with undesired processing of noise-free image pixels, which results in edge and texture blurring. So in the case of impulse noise removal, the aim of optimal filtering is to design noise reduction algorithms that would affect only corrupted image pixels, whereas the undistorted image pixels should be invariant under the filtering operation. Thus, in this thesis, a simple but effective impulse detector is presented to find out all pixels that are considered as noise before filtering. If a pixel is considered as noise, it will be replaced by a pixel in the neighborhood which is noise-free during filtering procedure later; otherwise, it is unchanged. So the performance of the impulse detector is crucial in the filtering because the whole filtering procedure depends on it. Simulation results show that the simple impulse detector has effective performance in detecting the noisy pixels in color images.(2) On the basis of the efficient and yet effective impulse noise detector, we propose a series of improved modified conventional filters. The conventional filters include the vector median filter (VMF), the basic vector directional filter (BVDF), the directional distance filter (DDF), and the self-adaptive algorithm (SAA) based on similarities.(3) Based on an integration of the simple impulse noise detector and a robust neuro-fuzzy network (RNFN), an effective impulse noise filtering method for color images is presented. It consists of two modes of operation, namely, training and testing (filtering). During training, the impulse detector is used to locate the noisy pixels in the color images for optimizing the RNFN. During testing, if a pixel is detected as a corrupted one according to the impulse detector, the trained RNFN will be triggered to output a new pixel to replace it. The proposed impulse noise filter is distinguished by the use of a novel membership function in the design of the adaptive RNFN, making the network robust to impulse noise.(4) Medical images are often corrupted by additive Gaussian noise. As to the filtering of medical images, it is very important to preserve edges and details. A major drawback of filtering is that it often blurs important structures along with noise. Scale-based filtering methods of scalar images have been studied in recent years. In this thesis, we generalize the scale-based filtering method from scalar images to vectorial images. Here we introduce three vectorial scale-based image filtering methods on the basis of conventional VMF, BVDF, and DDF. These new methods use local structure size or"object scale"information to arrest smoothing around fine structures. The object scale allows us to better control the filtering process by constraining smoothing in regions with fine details while permitting effective smoothing in the interior of homogeneous regions.As demonstrated by the extensive experimental results, all the proposed filtering methods compare favorably with conventional techniques in the capabilities of noise attenuation and details preservation, in both quantitative and qualitative measures.