Human Visual System for Complexity Reduction of Image and Video Restoration Vittoria Bruni 1 , Daniela De Canditiis 2 , and Domenico Vitulano 2 1 Dept. of SBAI, Faculty of Engineering, University of Rome ”La Sapienza” bruni@dmmm.uniroma1.it 2 Istituto per le Applicazioni del Calcolo ”M. Picone” (CNR) - Rome d.decanditiis-d.vitulano@iac.cnr.it Abstract. This paper focuses on the use of Human Visual System (HVS) rules for reducing the complexity of image and video restoration algorithms. Specifically, a fast HVS based block classification is proposed for distinguishing image blocks where restoration is necessary from the ones where it is useless. Some experimental results on standard test im- ages and video sequences show the capability of the proposed method in reducing the computing time of de-noising algorithms, preserving the visual quality of the restored sequences. Keywords: Human Visual System, Block classification, Complexity Re- duction, Image and Video Restoration. 1 Introduction A wide literature has been dedicated to image and video restoration with par- ticular attention to both quality and computational effort. In particular, the latter is fundamental for real time applications and for codecs transportability on common devices. Even though the more recent literature has taken a great advantage of using Human Visual System (HVS) mechanisms for improving cod- ing performance [1,2], for guiding image enhancement [3,4] or for detecting image anomalies [5], to the best of authors knowledge, the benefit from using HVS rules in restoration schemes for computational purposes has not yet been investigated. This paper aims at presenting a fast HVS-based blocks classification to be embedded into any de-noising algorithm in order to reduce its computing time. It is related to the Structural SIMilarity index (SSIM) [6], that is used for eval- uating the visual difference between two images. The proposed classification actually aims at distinguishing between: i) blocks where both noise and motion are perceived, ii) blocks where noise is perceptible while motion is not and iii) blocks where human eye is insensitive to both noise and motion. The goal is to adapt the restoration process to each block, according to its visible content. In particular, de-noising is inhibited if noise is imperceptible, while motion vector is not estimated if motion is not perceived. The computational gain depends on the processed frames and the selected restoration scheme. The larger the number of blocks where operations are inhibited and the more negligible the A. Berciano et al. (Eds.): CAIP 2011, Part II, LNCS 6855, pp. 261–268, 2011. c  Springer-Verlag Berlin Heidelberg 2011