Catálogo de publicaciones - libros

Image hash function based on the image content has applications in watermarking, authentication and image retrieval. This paper presents an algorithm for generating an image hash that is robust against content-preserving modifications and at the same time, is capable of detecting malicious tampering. Robust features are first extracted from the discrete wavelet transform followed by the Radon transform. Probabilistic quantization is then used to map the feature values to a binary sequence. Results show that the proposed method can resist perceptually insignificant modifications such as compression, filtering, scaling and rotation. It is also able to successfully detect content changing attacks such as insertion of foreign objects.

There are many applications, for example image copyright protection, where transformed images of a given test image need to be identified. The solution to the identification problem consists of two main stages. In stage one, certain representative features are detected for all images. In stage two, the representative features of the test image and the stored images are compared to identify the transformed images for the test image. We have reported the technique to extract robust representative features – corners – in our previous work [1]. This paper will focus on our stage-two work on effective corner matching technique for transformed image identification. Experimental results show that the proposed corner matching technique is very much effective in identifying the transformed images for a given test image.

In this paper, an automatic approach for detecting and extracting panels in a color comic image is proposed. Panel extraction is challenging because the background color, the background pixel locations, the panel shapes and the panel layout are not known in advance. In our approach, uniform color stripes are first identified and used as separators to segment the color comic page image into sub-regions in a recursive manner. Panels are recognized as the sub-regions that cannot be further segmented. The structure of the panels is thus obtained in the extraction process and it contains the layout of the panels as well as the reading order. Panel extraction is useful because: 1) the extracted panels can be better fitted into a handheld device for viewing; and 2) the panels can then be further analyzed to extract features used for content based indexing and retrieval.

Objective image quality assessment (QA), which automatically evaluates the image quality consistently with human perception, is essentially important for numerous image and video processing applications. We propose a new objective QA method for full reference model based on the energy of structural distortion (ESD). Firstly, we collect the characteristics of the structural information by the normalization processing for the reference image. Secondly, the information of ESD is gained by projecting the image onto the characteristic signal of the structural information independently. Finally the objective quality score is obtained by computing the differences of ESD between the reference and distorted images. In this paper, we propose one implementation with simple parameters for our image QA. Experimental results show that the proposed method is well consistent with the subjective quality score.

We propose a novel algorithm for the de-speckling of SAR images which exploits a priori statistical information from both the spatial and wavelet domains. In the spatial domain, we apply the Method-of-Log-Cumulants (MoLC), which is based on Mellin transform, in order to locally estimate parameters corresponding to an assumed Generalized Gaussian Rayleigh (GGR) model for the image. We then compute classical cumulants for the image and speckle models and relate them into their wavelet domain counterparts. Using wavelet cumulants, we separately derive parameters corresponding to an assumed generalized Gaussian (GG) model for the image and noise wavelet coefficients. Finally, we feed the resulting parameters into a Bayesian maximum a priori (MAP) estimator, which is applied to the wavelet coefficients of the log-transformed SAR image. Our proposed method outperforms several recently proposed de-speckling techniques both visually and in terms of different objective measures.

Spatially Squeezed Surround Audio Coding (SAC) was introduced as an approach to multi-channel audio compression which specifically aims to preserve original source localization information. In this paper, extensions to SAC that allow for the accurate coding of independent spatial sources overlapped in both frequency and time are described; these use compact side information. An evaluation of the coder applied to tone and band-pass spatial sources shows that SAC offers improved source localization performance while maintaining bit-rates, when compared with other state-of-the-art spatial audio coders.

A multi-resolution model often costs more storage space, its communications from the CPU to the graphics system is the bottleneck of the visualization process. In this paper, a multi-resolution mesh and a primitive are proposed. The primitive is used both in the storage stage and in the rendering stage, decreasing the storage size of model and the transmission amount of vertices to the graphics system. The efficiency is measured by means of tests and results compared with the previous, obtaining better storage space cost and transmission cost.

This paper proposes a cost-effective VLSI architecture to improve the three-step search (TSS) algorithm for efficient motion estimation. A weighted SAD is defined as the new distortion measure instead of SAD for motion vector selection to remedy the fault of the TSS algorithm. The proposed TSS architecture is superior to conventional TSS architecture in terms of coding performance. Moreover, the additional hardware implementation cost of the proposed architecture is relatively negligible. The proposed architecture achieves best tradeoff in terms of speed and hardware cost.