Catálogo de publicaciones - libros

This paper presents a novel method for annotating videos taken from the TRECVID 2005 data using only static visual features and metadata of still image frames. The method is designed to provide the user with annotation or tagging tools to incorporate multimedia data such as video or still images as well as text into searching or other combined applications running either on the web or on other networks. It mainly uses MPEG-7-based visual features and metadata of prototype images and allows the user to select either a prototype or a training set. It also adaptively adjusts the weights of the visual features the user finds most adequate to bridge the semantic gap. The user can also detect relevant regions in video frames by using a self-developed segmentation tool and can carry out region-based annotation with the same video frame set. The method provides satisfactory results even when the annotations of the TRECVID 2005 video data greatly vary considering the semantic level of concepts. It is simple and fast, using a very small set of training data and little or no user intervention. It also has the advantage that it can be applied to any combination of visual and textual features.

An important issue in embedding watermark bits in compressed video stream is to keep the bit-rate unchanged after the watermarking. This is a very difficult problem for high efficient compression methods such as H.264/AVC, because just one bit alteration in highly compressed bit-stream may widely affect the video content. In this paper we solve this problem by embedding watermark bit to the sign bit of the in Context Adaptive Variable Length Coding (CAVLC) of H.264/AVC. The algorithm yields no bit-rate change after the data hiding. Also, we can easily balance between the capacity of the watermark bits and the fidelity of the video. The simplicity of the proposed algorithm is an added bonus for the real-time applications. Our experiments show that the PSNRs of the video sequences after the data hiding are higher than 43dB.

A reduced uneven multi-hexagon-grid search algorithm for fast integer pel motion estimation in H.264/AVC is presented. The objective is to reduce the number of candidates for the block matching by predicting the likely area in which the minimum sum of absolute differences (SAD) can be taken. For this purpose, the proposed algorithm employs directionally partial hexagon search patterns utilizing the motion vectors computed in previous stages which supply the spatial correlation characteristics between adjacent macro blocks and the temporal ones between video frames. Experimental results show that the proposed method can save 39%~69% of computational complexity compared with the original one at the cost of negligible degradation on RD performance.

This paper proposes a lossless data hiding technique for JPEG images based on histogram pairs. It embeds data into the JPEG quantized 8x8 block DCT coefficients and can achieve good performance in terms of PSNR versus payload through manipulating histogram pairs with optimum threshold and optimum region of the JPEG DCT coefficients. It can obtain higher payload than the prior arts. In addition, the increase of JPEG file size after data embedding remains unnoticeable. These have been verified by our extensive experiments.

In this paper we introduce a fractal-based method over (complex-valued) Fourier transforms of functions with compact support . This method of “iterated Fourier transform systems” (IFTS) has a natural mathematical connection with the fractal-based method of “iterated function systems with greyscale maps” (IFSM) in the spatial domain [6,7]. A major motivation for our formulation is the problem of resolution enhancement of band-limited magnetic resonance images. In an attempt to minimize sampling/transform artifacts, it is our desire to work directly with the raw frequency data provided by an MR imager as much as possible before ‘“returning” to the spatial domain. In this paper, we show that our fractal-based IFTS method can be tailored to perform frequency extrapolation.

Reconstructed images in low bit-rate video coding often have visually annoying artifacts, such as blocking artifacts and corner outliers, usually due to block-based quantization and motion-compensated prediction. In this paper, we propose an adaptive deblocking algorithm based on local image characteristics. We remove ringing artifacts using a simple low-pass filter in the preprocessing. Among four filtering modes, an appropriate filter is selected adaptively, removing noticeable blocking artifacts in flat areas, while preserving details without blurring in highly complex areas. The proposed filter contains a non-symmetric filter a corner outlier removal filter. Simulation results show the effectiveness of the proposed deblocking algorithm in objective and subjective quality compared with those of the MPEG-4 deblocking filter.

‘’ is a ubiquitous strategy employed in the sensory information process system of mammals. Such strategy aims to find a representation of data in which the components of the representation are only rarely significantly active. This paper presents a and demonstrates that it may be used to separate natural images into different ingredients based on the sparse coding strategy. In such model an overcomplete dictionary is constructed by combining different type of complete or over-complete systems that can respectively deal with different image ingredients. Based on a sparse prior restriction, decomposition coefficients are inferred by maximizing a posterior distribution. The resulting coefficients belonging to different systems correspond to different image ingredients. The proposed provides a flexible framework for image ingredient separation which allows one to extract image structure of special interest.

An algorithm for lossy compression of vector data (vector maps, vector graphics, contours of shapes) was developed. The algorithm is based on optimal polygonal approximation for error measure and dynamic quantization of the vector data. The algorithm includes optimal distribution of the approximation line segments among the vector objects, optimal polygonal approximation of the objects with dynamic quantization and construction of the optimal variable-rate vector quantizer. The developed algorithm can be used for lossy compression of one-dimensional signals and multidimensional vector data.

In this paper, we introduce a new watermarking algorithm to embed an invisible watermark into the intra-frames of an MPEG video sequence. Unlike previous methods where each video frame is marked separately, our proposed technique uses high-order tensor decomposition of videos. The key idea behind our approach is to represent a fixed number of the intra-frames as a 3D tensor with two dimensions in space and one dimension in time. Then we modify the singular values of the 3D tensor, which have a good stability and represent the video properties. The main attractive features of this approach are simplicity and robustness. The experimental results show the robustness of the proposed scheme against the most common attacks including geometric transformations, adaptive random noise, low pass filtering, histogram equalization, frame dropping, frame swapping, and frame averaging.

Fractal image coding is a relatively recent technique based on the representation of an image by a map of self-similarities. In last years, most researchers focused their attention on the problem of speeding up the fractal coding process, while paying little attention to possible improvements of the objective and subjective image quality. In this paper, we investigate image quality measures, which could represent a reasonable alternative to the RMSE when finding a suitable map of similarities. Subjective assessments have been performed in order to compare performances of the selected quality metrics. Experimental results bear witness to the superiority of such a quality metric based on Fourier coefficients.