Real-time 2D–3D filtering using order statistics based algorithms

Abstract

The paper presents a review of the author's own results obtained in the last several years. Some examples of real-time processing of 2D and 3D images are described. In particular, we discuss the noise model and objective criteria that can be applied to characterize the performance of the processing algorithms. Several proposed algorithms based on RM approach are compared with other known ones, demonstrating the advantages in noise suppressing and preservation of fine image details and edges. A number of 2D and 3D image denoising filters are implemented on DSP, realizing real-time mode in the image processing. The performances of the proposed processing algorithms and the known ones are discussed and evaluated here.

Adaptive rank weighted switching filter for impulsive noise removal in color images

“…The choice of the dissimilarity measure strongly influences the properties of the resulting filter. Usually the angle and distance between vectors is utilized; however, various combinations of the magnitude and directional processing can also be applied [19][20][21][22][23][24][25][26][27][28]. Many filtering solutions define the vector x ð1Þ in (2) as their output, since vectors that diverge significantly from the samples of W appear in the higher indexed locations in their ordered sequence.…”

Multichannel and Multispectral Image Restoration Employing Fuzzy Theory and Directional Techniques

“…It is known that the principal difference between noise suppression in still images and video sequences, where information from previous and future frames may also be available, consists of finding the efficient use of several neighbour frames during processing, taking into account a possible motion between frames. In this chapter, a novel scheme to characterize the difference between pixels is proposed introducing gradients that are connected with pixel angular directions, and additionally, robust directional processing techniques presented in (Ponomaryov, 2007) (Ponomaryov & Gallegos et al, 2006). The gathering of such two methods realizes suppression of a noise, as well as preservation of fine image details on base on designed fuzzy rules and the membership degree of motion in a 3D sliding-window.…”

“…Important advantage of current filtering framework consists of using only two frames (past and present) reducing the processing requirements. We also realize the adaptation of several 2D algorithms in filtering of 3D video data: MF 3F, VGVDF (Trahanias & Venetsanopoulos, 1996), VVMF and VVDKNNVMF (Ponomaryov, 2007). Additionally, we have implemented the VKNNF, VATM (Zlokolica et al, 2006), and VAVDATM filters (Ponomaryov, 2007).…”

“…We also realize the adaptation of several 2D algorithms in filtering of 3D video data: MF 3F, VGVDF (Trahanias & Venetsanopoulos, 1996), VVMF and VVDKNNVMF (Ponomaryov, 2007). Additionally, we have implemented the VKNNF, VATM (Zlokolica et al, 2006), and VAVDATM filters (Ponomaryov, 2007). The video denoising can be realized in two forms: during temporal processing, or in spatiotemporal processing scheme (Zlokolica et al, 2005;Zlokolica et al, 2006), where additional 3D characteristics of a video sequence (motion and noise estimation) are researched and developed.…”

“…To justify the effectiveness of 3D technical proposals, the comparison with other filtering frameworks in video denoising were used (Zlokolica et al (2006); Ponomaryov & Gallegos et al (2006); Schulte & Huysmans et al (2006); Schulte & Witte et al (2006)). Reference filters: "Fuzzy Motion Recursive Spatio-Temporal Filter" (FMRSTF) (Zlokolica et al, 2006); an adaptation of FMRSTF employing only angles instead of gradients, named as "Fuzzy Vectorial Motion Recursive Spatio-Temporal Filter" (FVMRSTF); "Video Generalized Vectorial Directional Processing" (VGVDF) (Schulte & Huysmans et al, 2006), and also "Video Median M-type K-Nearest Neighbour" (VMMKNN) described in (Ponomaryov, 2007) were used as comparison in Gaussian denoising using numerous simulations and experiments. The algorithms MF_3F, VGVDF, VVMF, VVDKNNVMF, VKNNF, VATM, and VAVDATM were used as comparative ones to evaluate FCF -3D rendering.…”