237 Quality Assessment for Block-based Compressed Images and Videos with regard to Blockiness Artifacts * Donald Bailey 1 , Marco Carli 2, 3 , Mylene Farias 3 , Sanjit Mitra 3 1 Institute of Information Sciences and Technology, Massey University, New Zeeland D.G.Bailey@massey.ac.nz. 2 EE Dept., University of ROMA TRE, via della vasca navale 84, 00146 Roma, ITALY (Ph. +39.06.55177064 fax: +39 06 5579078, e-mail: carli@ele.uniroma3.it 3 ECE Dept., University of California Santa Barbara 93106 Santa Barbara, CA, USA (Ph. (805). 893.8312 fax: +(805) 893.3262, email: (mylene, mitra)@ece.ucsb.edu) * This research is supported in part by Italian National Research Council, in part by a National Foundation Grant No. CCR-0105404, in part by CAPES-Brazil, and by Microsoft Corp. ABSTRACT In this paper a novel quality metric for block-based compressed images and videos is proposed. The metric is designed to measure the “strength” of the blocking artifacts. The method is fast and robust. Experimental results show that there is a clear relationship between the strength of the detected blockiness and the compression ratio. The proposed algorithm is also able to distinguish between natural edges present in the original image, and blocking artifacts originated from compression. 1 INTRODUCTION Lossy image/video compression algorithms introduce several types of distortions into an image, such as, blockiness, blurring, mosquito effects, ringing, etc. [1] For low to moderate compression rates, there are very few visible artifacts and for most purposes the output image is perceptually indistinguishable from the original image. On the other hand, artifacts caused by high compression ratios are visible and may become highly annoying. Video communication is a very demanding application in terms of bit rate. State of the art video compression techniques enable the transmitted data rate to be reduced to few kilobits/second for low-resolution video conferencing making it affordable for a wireless system. These highly compressed video streams are consequently associated with the presence of highly annoying artifacts. Among the artifacts found in compressed images, blockiness is one of the most annoying ones. Block distortion or tiling is defined in ANSI T1.801.02-1996 as “Distortion of the image characterized by the appearance of an underlying block encoding structure.” Blockiness is a result of the DCT block-based compression technique used by some of the traditional compression algorithms, like for example JPEG or MPEG. These compression methods do not account for the correlation between neighbouring blocks, which causes the boundaries of each block to appear as discontinuities (edges) in the decompressed images. Figure 2 illustrates this type of artifact. Clearly, blockiness is not the only type of artifact present in this image. Two frequently used measures of the quality of images/videos are the mean absolute error (MSE) and the mean square error (PSNR) [2]. These measures are simple to calculate, since they are based solely on the differences between the degraded image/video and the reference. Nevertheless, the outputs of these measures do not always correlate well with human judgments of quality [3]. Besides, in most communication applications the reference image or video is not available and such measures cannot be computed. In order to provide users of communication systems with a quality of measure for compressed images or video, a fast and efficient non-reference metric is definitely needed. In this paper a new quality metric based on measure of the strength of the blocking artifacts [4] is proposed. It takes advantage of the fact that edges in blocking artifacts appear at regular intervals. Experimental results show that the proposed metric is fast and robust. Figure 1: “Barbara”.