OFFSET BASED LEAKY PREDICTION FOR ERROR RESILIENT ROI CODING Zhengyi Luo, Li Song and Shibao Zheng Institute of Image Communication and Information Processing, Shanghai Jiao Tong University {llzzyynjupt, song_li, sbzh}@sjtu.edu.cn ABSTRACT During the period of transmission, video data usually suffer from transmission errors inevitably. Intra update is a common approach to stop error propagation. However, damaged images cannot recover until next update in case of errors, which often leads to annoying effect. In this paper, we propose an enhanced leaky prediction approach that enables the Region-Of-Interest (ROI) of images to recover gently from the immediate succeeding frame of erroneous ones in favor of better human perception. Moreover, an optimized offset compensation technique is designed to improve coding performance. Experimental results show that the proposed scheme can achieve better image quality for ROI and the fluctuation of bitrate is greatly reduced, compared to the intra update method. Index Terms— error resilient, ROI, leaky prediction, offset compensation, H.264/AVC 1. INTRODUCTION With the rapid development of multimedia technologies, there is an increasing demand for video transmission nowadays. However, common underlying channels, such as the Internet or wireless networks, usually provide only best- effort services. Thereby video data often suffer from transmission errors inevitably. Research on Human Visual System (HVS) reveals that people generally pay more attention to the Region-Of- Interest (ROI) areas, such as the heads in fig. 1. So in order to achieve better visual quality, ROIs should be treated specially. As video frames are usually inter-frame coded, error propagation in ROI areas has to be properly dealt with. Error concealment can repair erroneous areas of images using temporal or spatial neighboring ones. But such recovery is limited for ROI areas with complex motion and textures. Thus, other error resilience tools are still desired. Some schemes are proposed to enhance the robustness of ROI areas against errors at the source encoder level. For example, [1] employs a nonlinear transform in ROI areas during pre- and post-processing, and [2] performs double motion estimation for ROIs. Besides, some research works [3~5] choose to combine the resilience feature of source encoder, such as Flexible Macroblock Ordering (FMO) tools, with transport level protection like Forward Error Correction (FEC). With these methods, Unequal Error Protection (UEP) can be applied to provide more protection to ROI areas than backgrounds during transmission. (a) (b) Fig. 1 ROI from (a) “Foreman”, (b) “Silent”. In this paper, we contribute another effective source level error resilient tool for enhancing the error robustness of ROI without sacrificing much coding efficiency. Our scheme designs a special weighted prediction approach for ROI areas. On one hand, leaky prediction technique is adopted to ensure erroneous regions recover gradually with time. On the other hand, an optimized offset compensation can effectively prevent rate-distortion performance from dropping and keep the constant quality of ROI when no error happens. Furthermore, this scheme also reduces fluctuation of bitrate significantly, which is friendly to packetization or FEC processing at the transport layer. It should be noted that our scheme needs no changes to the H.264/AVC standard [6] and can be easily implemented by existing weighted prediction tools. The remainder of this paper is organized as follows. In Section 2, the proposed scheme and its implementation are described in detail. Experimental results validating the effectiveness of our scheme are shown in Section 3. Section 4 draws the conclusion. 2. OFFSET BASED LEAKY PREDICTION Considering ROIs usually don’t take up too much space of the whole image, an intuitive method of stopping error propagation is periodical intra-update of ROI. However, such method often incurs annoying visual perception at low frame rate scenarios such as the mobile video phone, since quality of ROI may be always poor until the next intra- update. Another gentle way to combat error propagation is 145 978-1-4244-4291-1/09/$25.00 ©2009 IEEE ICME 2009