Yu Mei

A Digital Watermarking Algorithm Based on Region of Interest for 3D Image

A digital watermarking algorithm based on depth perceptual region of interest (DP-ROI) for copyright protection of three dimension (3D) images is proposed. First, DP-ROI is defined according to the three-dimensional depth sensation of human visual system and extracted through depth image and gray image. Then, gray image is transformed with discrete cosine transformation (DCT). Several middle frequency DCT coefficients in DP-ROI are adjusted with smaller strength to embed watermark. Finally, by computing the correlation of DCT middle frequency coefficients, and blind watermark is extracted. The experimental results show that the proposed watermarking algorithm has good robustness against image processing attacks, such as JPEG compression, Gaussian noise addition through selection of watermark embedding strength.

An efficient fast mode selection for intra prediction

In order to maximize the coding performance, H.264 adopts RDO (rate distortion optimization) at the cost of a very high computational complexity. A special focus is spent on fast intra prediction mode selection in this paper. To reduce complexity associated with intra prediction mode selection, an efficient fast mode selection for intra prediction is proposed. First, the Pans fast algorithm based on edge direction histogram is improved. Second, a new improved algorithm is joined together with a 44 intra prediction mode selection algorithm based on the feature of the reference pixels. Experimental results show that, compared with the H.264 encoder JM61, the proposed scheme reduces 58% coding time for I-frame while maintaining the PSNR and only increasing bit rate a little. Furthermore, compared with Pans scheme based on edged direction histogram, the proposed scheme reduces 24% coding time for I-frame while maintaining similar PSNR and bit rate.

New adaptive algorithm for video watermarking

In this paper, a new adaptive watermarking algorithm is presented for video image coding. Operating on the compressed video data-stream directly, we embed watermarking information into motion vectors. According to watermarking information and the feature vectors of motion vectors, the proposed algorithm can adaptively select the embedding scheme with the aim to get good watermarked image quality. The watermark is extracted blindly without the original images and experimental results show that the proposed algorithm has good compatibility with the present compression standards, and it is very simple and efficient.

Quality and performance evaluation of ray-space interpolation for free viewpoint video systems

Ray-space is the main technology to realize Free Viewpoint Video (FVV) systems with complicated scenes. One of the key technologies to make the ray-space feasible is interpolation. In this paper, two fast interpolation methods based on feature points and directionality detection are compared with block based matching method, and the experimental results show that the two methods improve visual quality as well as PSNRs of rendered virtual viewpoint image greatly while computational burden is also reduced significantly.

Multiple description scalable video coding based on 3D lifted wavelet transform

In this work, a new method to deal with the unconnected pixels in motion compensated temporal filtering (MCTF) is presented, which is designed to improve the performance of 3D lifted wavelet coding. Furthermore, multiple description scalable coding (MDSC) is investigated, and novel MDSC schemes based on 3D wavelet coding are proposed, using the lifting implementation of temporal filtering. The proposed MDSC schemes can avoid the mismatch problem in multiple description video coding, and have high scalability and robustness of video transmission. Experimental results showed that the proposed schemes are feasible and adequately effective.

New approach to reducing blocking effects in stereo video coding

The correlation property of blocks may be damaged in block-based image coding, and blocking effects can be induced. Some algorithms were developed for reducing the blocking effects. However, algorithms for reducing the blocking effects in video images, especially in stereo video images, should be simple and have low computational load. In this paper, causes of inducing blocking effects in stereo video coding are analyzed, and a new algorithm for reducing three types of blocking effects is presented. For the three types of blocking effects in I-frames, three filtering schemes are proposed. For the blocking effects in P-frames and B-frames, a new method of optimizing motion estimation and disparity estimation is proposed. Experimental results show that the proposed algorithm is simple and efficient.

New approach to multiple description pyramid lattice vector quantization for robust image coding

Multiple description coding (MDC) is a new approach to alleviate the problem of packet loss in image communication over erasure channels, such as Internet and wireless networks. With decomposing an image into several descriptions and transmitting them over different channels, MDC improves the reconstructed image quality with respect to losing portions of the packet. In this paper, a new image coding scheme, called as multiple description pyramid lattice vector quantization (MDPLVQ), is proposed for robust image coding, utilizing wavelet tree structure of image and two alterable quantizers to control the redundancy of MDPLVQ. Experimental results show that the presented MDPLVQ is quite efficient.

Generation of virtual intermediate views in object-based stereo video coding

A new approach to generating virtual intermediate views for stereo videoconferencing system is presented in this paper, including extraction of stereo video objects, estimation of disparity field, and generation of virtual intermediate views. A scheme of extraction of stereo video object is first given, an adaptive multi-level disparity estimation algorithm based on characteristic disparity vectors is proposed. Then, an arbitrary intermediate disparity field is obtained with disparity interpolation and the corresponding virtual intermediate view is generated. Experimental results show that the proposed method is efficient to generate virtual intermediate views.