Local Feature Regions Research Articles

A feature-based digital watermarking scheme for halftone image

It is a challenging work to design a robust halftone image watermarking scheme against desynchronization attacks. In this paper, we propose a feature-based digital watermarking method for halftone images with low computational complexity, good visual quality and reasonable resistance toward desynchronization attacks. Firstly, the feature points are extracted from host halftone image by using multi-scale Harris–Laplace detector, and the local feature regions (LFRs) are constructed according to the feature scale theory. Secondly, discrete Fourier transform (DFT) is performed on the LFRs, and the embedding positions (DFT coefficients) are selected adaptively according to the magnitude spectrum information. Finally, the digital watermark is embedded into the LFRs by quantizing the magnitudes of the selected DFT coefficients. By binding the watermark with the geometrically invariant halftone image features, the watermark detection can be done without synchronization error. Simulation results show that the proposed scheme is invisible and robust against common signals processing such as median filtering, sharpening, noise adding, and JPEG compression, etc., and desynchronization attacks such as rotation, scaling, translation (RST), cropping, local random bend, and print-scan, etc.

Invariant image watermarking using multi-scale Harris detector and wavelet moments

Desynchronization attack is known as one of the most difficult attacks to resist, which can desynchronize the location of the watermark and hence causes incorrect watermark detection. It is a challenging work to design a robust image watermarking scheme against desynchronization attacks. Based on multi-scale Harris detector and wavelet moment theory, we propose a new content based image watermarking algorithm with low computational complexity, good visual quality and reasonable resistance toward desynchronization attacks in this paper. Firstly, the steady image feature points are extracted from the origin host by using multi-scale Harris detector, and the local feature regions (LFRs) are constructed adaptively according to the feature scale theory. Then, the LFRs are image normalized, and significant regions are obtained from the normalized LFRs by utilizing the invariant centroid theory. Finally, the digital watermark is embedded into the LFRs by modifying wavelet moment invariants of the significant regions. By binding the watermark with the geometrically invariant image features, the watermark detection can be done without synchronization error. Experimental results show that the proposed image watermarking is not only invisible and robust against common image processing operations as sharpening, noise adding, and JPEG compression etc, but also robust against the desynchronization attacks such as rotation, translation, scaling, row or column removal, cropping, and local random bend etc.

Boosted cascade of scattered rectangle features for object detection

This paper presents a variant of Haar-like feature used in Viola and Jones detection framework, called scattered rectangle feature, based on the common-component analysis of local region feature. Three common components, feature filter, feature structure and feature form, are extracted without concerning the details of the studied region features, which cast a new light on region feature design for specific applications and requirements: modifying some component(s) of a feature for an improved one or combining different components of existing features for a new favorable one. Scattered rectangle feature follows the former way, extending the feature structure component of Haar-like feature out of the restriction of the geometry adjacency rule, which results in a richer representation that explores much more orientations other than horizontal, vertical and diagonal, as well as misaligned, detached and non-rectangle shape information that is unreachable to Haar-like feature. The training result of the two face detectors in the experiments illustrates the benefits of scattered rectangle feature empirically; the comparison of the ROC curves under a rigid and objective detection criterion on MIT+CMU upright face test set shows that the cascade based on scattered rectangle features outperforms that based on Haar-like features.

Local region-based image quality assessment independent of JPEG and JPEG2000 coded color images

The importance of the perceived quality measurement is fundamental for many image processing applications, such as compression, acquisition, restoration, enhancement, and reproduction. Color information is also of great importance for the perceived image quality, although perceived information is mainly represented by luminance. We present a computational and memory-efficient no-reference image quality assessment model independent of JPEG and JPEG2000 coded color images based on local regions. We also present the discrimination algorithm for these two types of coded images. The features of local regions are blockiness around the block boundary, average absolute difference between adjacent pixels within the block, and zero crossing rate within the block of the image. We validate the performance of our model on our subjective database, which shows good quality prediction performance, and the model's generalization ability is also verified on the other database.

Feature-based Digital Image Watermarking Scheme Robust to Geometric Attacks

摘要: 提出了一种抵抗一般性几何攻击的数字水印新算法. 该算法首先利用多尺度 Harris 检测算子从载体图像中提取出稳定的特征点; 然后根据特征尺度自适应确定局部特征区域, 并对其实施归一化处理; 最后结合预失真补偿理论, 采纳 DFT 中频幅值比较策略将水印信息重复嵌入到多个归一化局部特征区域内. 仿真实验表明, 本文算法不仅具有较好的透明性, 而且对常规信号处理和一般性几何攻击均具有较好的鲁棒性. 关键词: 图像水印 / 一般性几何攻击 / 特征点 / 归一化 / 预失真补偿

A feature-based robust digital image watermarking against geometric attacks

Based on scale space theory and an image normalization technique, a new feature-based image watermarking scheme robust to general geometric attacks is proposed in this paper. First, the Harris–Laplace detector is utilized to extract steady feature points from the host image; then, the local feature regions (LFR) are ascertained adaptively according to the characteristic scale theory, and they are normalized by an image normalization technique; finally, according to the predistortion compensation theory, several copies of the digital watermark are embedded into the nonoverlapped normalized LFR by comparing the DFT mid-frequency magnitudes. Experimental results show that the proposed scheme is not only invisible and robust against common signals processing methods such as median filtering, sharpening, noise adding, and JPEG compression etc., but also robust against the general geometric attacks such as rotation, translation, scaling, row or column removal, shearing, local geometric distortion and combination attacks etc.

Novel Intra Deinterlacing Algorithm Using Content Adaptive Interpolation

This paper presents a novel intra deinterlacing algorithm (NID) based on content adaptive interpolation. The NID consists of three steps: pre-processing, content classification, and content adaptive interpolation. There are also three main interpolation methods in our proposed NID, i.e. modified edge-based line averaging (M-ELA), gradient directed interpolation (GDI), and window matching method (WMM). Each proposed method shows different performances according to spatial local features. Therefore, we analyze the local region feature using the gradient detection and classify each missing pixel into four categories. And then, based on the classification result, a different de-interlacing algorithm is activated in order to obtain the best performance. Experimental results demonstrate that the NID method performs better than previous techniques.

Object based video retrieval with local region tracking

This paper describes a method for video retrieval system based on local invariant region descriptors. A novel framework is proposed for combined video segmentation, content extraction and retrieval. A similarity measure, previously proposed by the authors based on local region features, is used for video segmentation. The local regions are tracked throughout a shot and stable features are extracted. The conventional key frame method is replaced with these stable local features to characterise different shots. A grouping technique is introduced to combine these stable tracks into meaningful object clusters. The above method can handle the different scales of object appearance in videos. Compared to previous video retrieval approaches, the proposed method is highly robust to camera and object motions and can withstand severe illumination changes. The proposed framework is applied to scene and object retrieval experiments and significant improvement in performance is demonstrated.

Component-based subspace linear discriminant analysis method for face recognition with one training sample

Many face recognition algorithms/systems have been developed in the last decade and excellent performances have also been reported when there is a sufficient number of representative training samples. In many real-life applications such as passport identification, only one well-controlled frontal sample image is available for training. Under this situation, the performance of existing algorithms will degrade dramatically or may not even be implemented. We propose a component-based linear discriminant analysis (LDA) method to solve the one training sample problem. The basic idea of the proposed method is to construct local facial feature component bunches by moving each local feature region in four directions. In this way, we not only generate more samples with lower dimension than the original image, but also consider the face detection localization error while training. After that, we propose a subspace LDA method, which is tailor-made for a small number of training samples, for the local feature projection to maximize the discrimination power. Theoretical analysis and experiment results show that our proposed subspace LDA is efficient and overcomes the limitations in existing LDA methods. Finally, we combine the contributions of each local component bunch with a weighted combination scheme to draw the recognition decision. A FERET database is used for evaluating the proposed method and results are encouraging.

Localisation in large-scale environments

This paper describes a localisation framework that combines the accuracy of feature maps with the scalability of topological maps. The map is structured as a graph of nodes where each node defines a local region feature map. This breaks the localisation process into a combination of regional feature tracking and node-to-node context switching. As part of the practical implementation of the localisation system, we introduce a batch data association method that uses the simultaneous observation of multiple features to determine data associations in a manner decoupled from the vehicle pose estimate. We also present an observation-based dead reckoning procedure that estimates vehicle motion in place of odometry and does not require a kinematic vehicle model. Experimental results demonstrate that this approach is capable of localising in large-scale outdoor environments. We perform tests in an inner city park and a suburban street using a scanning range laser as the sole information source. The diverse nature of these two environments indicates that these techniques have broad application.

Protein solution structure determination using distances from two-dimensional nuclear Overhauser effect experiments: effect of approximations on the accuracy of derived structures.

Solution structures for many proteins have been determined to date utilizing interproton distance constraints estimated from two-dimensional nuclear Overhauser effect (2D NOE) spectra. Although the simple isolated spin pair approximation (ISPA) generally used can result in systematic errors in distances, the large number of constraints enables protein structure to be defined with reasonably high resolution. Effects of these systematic errors on the resulting protein structure are examined. Iterative relaxation matrix calculations, which account for dipolar interactions between all protons in a molecule, can accurately determine internuclear distances with little or no a priori knowledge of the molecular structure. The value of this additional complexity is also addressed. To assess these distance determination methods, hypothetical "experimental" data, including random noise and peak overlap, are calculated for an arbitrary "true" protein structure. Three methods of obtaining distance constraints from 2D NOE peak intensities are examined: one entails a conservative use of ISPA, one assumes the ISPA to be fairly accurate, and one utilizes an iterative relaxation matrix method called MARDIGRAS (matrix analysis of relaxation for discerning the geometry of an aqueous structure), developed in this laboratory. A distance geometry algorithm was used to generate a family of structures for each distance set. The quality of the average structure from each family was good. The root-mean-square deviation of that average structure from the true structure was improved about 2-5% using the more restrictive rather than the more conservative ISPA approach. Use of MARDIGRAS in a conservative fashion--i.e., with a poor initial model--resulted in improvement in the root-mean-square deviation by 8-15%. With a better initial model, MARDIGRAS obtained even more accurate distances. MARDIGRAS also permits analysis of 2D NOE data at longer mixing times, yielding additional distances. Use of more restrictive ISPA distances did, however, result in a few systematically incorrect structural features in local regions of the protein, producing distortions of 2-3 A. Comparison between experimental data and spectra calculated for the structures correlates with root-mean-square deviation, offering a method of structure evaluation. An R factor for evaluating fit between experimental and calculated 2D NOE intensities is proposed.