Fig 4 - uploaded by George Tzagkarakis
Content may be subject to copyright.
Histogram of the estimated values for the characteristic exponent, α, for the set of 208 texture images of size 128 × 128.
Source publication
This paper addresses the construction of a novel efficient rotation-invariant texture retrieval method that is based on the alignment in angle of signatures obtained via a steerable sub-Gaussian model. In our proposed scheme, we first construct a steerable multivariate sub-Gaussian model, where the fractional lower-order moments of a given image ar...
Context in source publication
Context 1
... histogram of the estimated characteristic exponent values for the 208 textures is shown in Fig. 4. We observe that only 28% of the textures exhibit very strong Gaussian statistics, corresponding to α values equal to 2, which belong mainly to pyramid subband coefficients of images 4, 5 and 9. Table II shows the average value of α, over all the pyramid subbands, for each texture class. We observe that images 4, 5 and 9 are exactly ...
Similar publications
In this paper, we propose a novel approach for intensity based atlas construction from a population of anatomical images, that estimates not only a template representative image but also a common optimal parameterization of the anatomical variations evident in the population. First, we introduce a discrete parameterization of large diffeomorphic de...
We propose a learning method which introduces explicit knowledge to the object correspondence problem. Our approach uses an a priori learning set to compute a dense correspondence field between two objects, where the characteristics of the field bear close resemblance to those in the learning set. We introduce a new local shape measure we call the...
Real-world phenomena involve complex interactions between multiple signal modalities. As a consequence, humans are used to integrate at each instant perceptions from all their senses in order to enrich their understanding of the surrounding world. This paradigm can be also extremely useful in many signal processing and computer vision problems invo...
In in this paper, we propose a new model regarding foreground and shadow detection in video sequences. The model works without detailed a priori object-shape information, and it is also appropriate for low and unstable frame rate video sources. Contribution is presented in three key issues: 1) we propose a novel adaptive shadow model, and show the...
Group studies of functional magnetic resonance imaging datasets are usually based on the computation of the mean signal across subjects at each voxel (random effects analyses), assuming that all subjects have been set in the same anatomical space (normalization). Although this approach allows for a correct specificity (rate of false detections), it...
Citations
... This new method provides more information about the edge and direction of texture images compared to previous tensor product wavelets (such as db wavelets). Tagarakis et al. [21] proposed a method for non-rotational tissue retrieval that uses the non-Gaussian behavior of sub band coefficient distributions and shows tissue information with a guided pyramid. Hahn et al. [22] proposed immutable Gabor representations in which each representation requires only a few summaries of Gabor filter impact responses; Texture properties are extracted from these new representations to retrieve the unchanged texture image. ...
In the next decade, machine vision technology will have an enormous impact on industrial works because of the latest technological advances in this field. These advances are so significant that the use of this technology is now essential. Machine vision is the process of using a wide range of technologies and methods in providing automated inspections in an industrial setting based on imaging, process control, and robot guidance. One of the applications of machine vision is to diagnose traffic accidents. Moreover, car vision is utilized for detecting the amount of damage to vehicles during traffic accidents. In this article, using image processing and machine learning techniques, a new method is presented to improve the accuracy of detecting damaged areas in traffic accidents. Evaluating the proposed method and comparing it with previous works showed that the proposed method is more accurate in identifying damaged areas and it has a shorter execution time.
... This new method provides more information about the edge and direction of texture images compared to previous tensor product wavelets (such as db wavelets). Tagarakis et al. [21] proposed a method for non-rotational tissue retrieval that uses the non-Gaussian behavior of subband coefficient distributions and shows tissue information with a guided pyramid. Hahn et al. [22] proposed immutable Gabor representations in which each representation requires only a few summaries of Gabor filter impact responses; Texture properties are extracted from these new representations to retrieve the unchanged texture image. ...
Machine vision technology will have a significant impact on all industrial works in the next decade due to the latest technological advances in this field, and these advances are such that the use of this technology is now vital. Machine vision involves the process of applying a wide range of technologies and methods in providing automated inspection based on imaging, process control, and robot guidance in industrial applications. One of the applications of car vision is to diagnose traffic accidents. Moreover, car vision is utilized for detecting the amount of damage to vehicles during traffic accidents. In this article, using image processing and machine learning techniques, a new method is presented to improve the accuracy of detecting damaged areas in traffic accidents. Evaluating the proposed method and comparing it with previous works showed that the proposed method is more accurate in identifying damaged areas and it has a shorter execution time.
... Alpha-stable distribution with heavy tails has been used in the field of image processing and pattern recognition. George Tzagkarakis et al. (2008) developed the rotation invariant texture retrieval approach, in which a steerable multivariate sub-Gaussian model is derived from alpha-stable distribution and the covariance between the orientation subbands of the corresponding steerable pyramid of wavelet transform at the same or at the adjacent decomposition levels are calculated from the covariations, and then the similarity is performed using a matrix-based norm between covariance. When the probability density of texture information is not symmetric, the skewed alpha-stable distribution is a more reasonable tool to model texture. ...
... Texture retrieval systems can be implemented in the pixel-domain (Basso et al. 2011) or the transform-domain (Tzagkarakis et al. 2008). Tzagkarakis et al. (2008) have pointed out that the structure of coefficients resulted from a linear invertible transform is simpler to model than the structure of the original image in the pixel-domain. ...
... Texture retrieval systems can be implemented in the pixel-domain (Basso et al. 2011) or the transform-domain (Tzagkarakis et al. 2008). Tzagkarakis et al. (2008) have pointed out that the structure of coefficients resulted from a linear invertible transform is simpler to model than the structure of the original image in the pixel-domain. The twodimensional wavelet transform is a powerful tool to model the edges, ridges, and lines in real-world images (Mallat 1998). ...
This paper proposed an efficient texture retrieval method for indexing images with heavy-tailed distribution, such as biomedical images, sonar images and natural texture images. In the proposed scheme, a multivariate Log-Gaussian mixture model (MLGMM) was used to model the sharp peaks, heavy tails, and even the multimodal statistical properties of two-dimension Gabor coefficients of texture under different scales and orientations. The parameters of MLGMM are estimated by expectation maximum (EM). In our scheme, each class of texture is modeled by one MLGMM and Bayesian classification is implemented by feeding the output of MMLGM into the Bayesian classifier. Experiments on feature extraction and similarity measurement have been done to demonstrate the effectiveness of our proposed algorithm. Extensive experiments have validated that our retrieval scheme has an average retrieval rate of 2% higher than other related texture statistical techniques.
... In [70], an approach has been suggested for non-rotating textures retrieval in which non-Gaussian behavior of the distribution of the sub-band coefficients is applied, and texture information is represented with a manageable pyramid. In [24], the non-variable Gabor representations has been suggested in which each representation requires just a few summations of Gabor filter impulse response from which the texture features are extracted for non-variable texture images. ...
Recently, we have been witnessing a tremendous rise in digital image quantities, which in return calls for an adjustment and management system to fulfill user’s queries in the shortest time with maximum accuracy. In this regard, Content-Based Image Retrieval (CBIR) approaches have gained unprecedented attention. In CBIR systems, image search is based on their actual contents instead of textual annotations.
Due to the fact that users do not think of low-level image features such as color, texture, structure, and shape and are looking for high-level image features or semantic features while querying images, the performance of image retrieval systems becomes weak. On the one hand, the huge amount of extracted features and the complexity of feature spaces are considered as two main challenges in image retrieval study area. Therefore, this article trying to extract the key features of the image in order to increase the accuracy and speed of image recovery over big data.
This study combines two feature extraction techniques namely Census Transform Histogram (CENTRIST) and Rotated Local Binary Pattern (RLBP) following by Kernel Principal Component Analysis (PCA) method to reduce the dimensional feature space. In fact, after feature extraction phase we utilize Adaboost M2 classifying method on the train data to learn the different classes of images that are existed in the database. Furthermore, instead of using RGB color space, images are transformed to HSV color space. The reason for using the HSV color space instead of the RGB one is the fact that it is closer to human perception.
Performance evaluations of the proposed method are conducted on the Corel-1 K and UW datasets. Simulation results indicate that proposed method performs better than other methods.
... Based on this sub-space decomposition, energy features have been used as a signature that represents the salient texture attributes for texture indexing [32]. For the SM stage, numerous approaches have been proposed to compare images based on their feature representations using perceptual studies, heuristics and statistical principles [3]- [6], [9], [11], [14], [19], [33]- [35]. For statistical principles, we highlight the seminal work of Do and Vetterli [4] and its numerous extensions [12], [16], [19], [33]. ...
... Examples, include the modeling of the inter-band dependencies [10], [11], [13], [16], [21], and the selection of more complex representations using Curvelet, frames, and other tools from signal processing. [3], [5]- [9], [11], [14], [15], [17], [18], [20], [21], [36]. ...
... Each texture of the Full STex, Full ALOT, Reduced STex, and Reduced ALOT is divided into 16 non-overlapping textures of 256 × 256 pixels. After these divisions, the color images are transformed into grayscale versions and, subsequently, all images in the collections are normalized to zero mean and unit variance, which is a standard normalization used in previous studies [4], [5], [8], [9], [63]. In total, we use six databases with different image sizes and number of textures, which offer a rich context to evaluate the potential of WP based texture indexing. ...
Wavelet Packets (WPs) bases are explored seeking new discriminative features for texture indexing. The task of WP feature design is formulated as a learning decision problem by selecting the filter-bank structure of a basis (within a WPs family) that offers an optimal balance between estimation and approximation errors. To address this problem, a computationally efficient algorithm is adopted that uses the tree-structure of the WPs collection and the Kullback-Leibler divergence as a discrimination criterion. The adaptive nature of the proposed solution is demonstrated in synthetic and real data scenarios. With synthetic data, we demonstrate that the proposed features can identify discriminative bands, which is not possible with standard wavelet decomposition. With data with real textures, we show performance improvements with respect to the conventional Wavelet-based decomposition used under the same conditions and model assumptions.
... The rotation invariance is one of the important properties of CBIR method. In order to achieve high quality, rotation invariant texture image retrieval, Tzagkarakis has developed (Tzagkarakis et al. 2008) a novel, efficient and rotation-invariant texture-image-retrieval method, which is based on the alignment in angle of signatures obtained through a steerable sub-Gaussian model. The feature extraction step consists of estimating the covariations between the orientation sub-bands of the corresponding steerable pyramid at the same, or at adjacent decomposition levels. ...
This paper proposes wavelet-based orthogonal polynomial model for content-based image retrieval. The wavelet coefficients are categorised into low-frequency and high-frequency based on criteria. The criteria are adaptively determined and fixed according to the nature and structure of the image, because the wavelet-based orthogonal polynomial model spatially localises the frequency information of image. Wavelet technique is employed to derive a multiresolution pyramid structure. Features such as spectrum of energy and spatial relation among the pixels are extracted at optimum level that helps in the formation of a feature vector. The Bhattacharyya distance and orthogonal cosine similarity method is employed to find the distance value between the feature vectors of the query and target images. The proposed method is simple and suitable for fast response requirements since the features extracted at the optimum-level image contain only fewer dominant wavelet coefficients. The efficiency of the proposed feature vector is experimented with standard image databases. Also, the proposed system is compared with the other existing methods.
... The steerable transform has been successfully employed by Do et al. [6] to extract rotation-invariant features and hidden Markov field model (HMM) to model dependencies between coefficients. Tzagkarakis et al. [7] also used the SP to get a rotation-invariant features through a multivariate α-stable as a special case of the Gaussian Scale Mixture model. ...
... Then, the minimum divergence corresponding to the estimated rotation angle is taken after an exhaustive optimization process. Thus, the divergence between the image query Q and the target image I becomes a minimization problem of the Jeffrey divergence (7). In what follows, the elements annotated by Q belong to the image query features, while the ones annotated by I belong to the target image features. ...
In this paper, we address the problem of rotation invariance in the context of texture retrieval. For this, we propose a framework based on the well-known copula theory which is considered one of the most powerful statistical tools. Prior to apply a such model, we first use the steerable pyramid SP as one of the most relevant transforms. Then, we build a steerable Gaussian copula model which offers a good fitting of the SP coefficients distribution while taking into consideration their rotation invariance property. Finally, we derive a closed-form of the Jefferey divergence as a similarity measure. The latter consists on an angular alignment between the query and the target texture features. Experiments have been conducted on USC database, good performances in term of retrieval rates are achieved compared to previously proposed copula models.
... The FLOM algorithm, also called the least mean p-norm (LMP) algorithm in the adaptive filtering literature [13,14], has been shown to be effective in many non-Gaussian interference environments. It has been applied to seismic system identification, synthetic aperture radar (SAR), magnetoencephalography, and medical ultrasonic imaging, etc., where Bernoulli-Gaussian models [15,16] or symmetrical alpha-stable models [13,14,17] are often used for impulsive interference with an algebraic tail. It has been shown that the FLOM algorithm performs significantly better than the conventional LMS (2nd-order norm) algorithm. ...
The normalized fractionally-lower order moment (NFLOM) algorithm differs from the normalized least mean square (NLMS) algorithm in that it minimizes the lower order moment (p2) of the error rather than the variance (p=2). This paper first evaluates the performances of the NFLOM for space-time adaptive processing in heavy-tailed compound K clutters in terms of the excess mean square error (MSE), misalignment, beampatterns, and output signal-to-interference-and-noise-ratio (SINR). The results show that the MSE curve of a small-order NFLOM exhibits faster convergence but higher steady-state error than a large-order NFLOM. Second, this paper proposes a new variable-order FLOM algorithm to dynamically change the order during adaptation, thus achieving both fast initial convergence and low steady-state error. The new algorithm is applied to STAP for Gaussian and non-Gaussian clutter suppression. The simulation results show that it achieves the best compromise between fast convergence and low steady-state error in both types of clutters.
... In addition, distributions of sub-Gaussian processes at any time t are symmetric; even more, if X α is multi-dimensional and the components of G are independent of each other, then the distribution of X α (t) is rotation-invariant. These properties made sub-Gaussian processes not only important theoretically, but also very popular as stochastic models in many areas including signal processing [7], mathematical finance [8] and image processing [12]. ...
In this paper, we first prove that one-parameter standard α-stable sub-Gaussian processes can be approximated by processes constructed by integrals based on the Poisson process with random intensity. Then we extend this result to the two-parameter processes. At last, we consider the approximation of the subordinated fractional Brownian motion.
... While estimating model parameters may require substantial computational load for algorithm training, the KL-similarity measurement is approximated with a fast algorithm based on a Monte Carlo method that is, comparable to the computation of the Euclidean distance between two feature vectors. Another alternative for statistical modeling has been recently presented in [20]. A multivariate sub-Gaussian model is tied up to the sub-band coefficients of a steerable pyramid providing a statistical description of the non-Gaussian nature of coefficient distribution in addition to the interactions and inter-dependencies across scales and orientations. ...
... Yes [20] GWT Multivariate Wald-Wolfowitz test mean+variance (2/sub-band) Yes [32] DT-CWT (L scale levels) Weibull-KL scale+shape (2/sub-band) No [38] Gamma-KL scale+shape (2/sub-band) No [38] Copula-KL (RGB) 2 Weibull param./sub-band + definite matrix No [39] CT (L scale levels) GGD-KL scale+shape (2/sub-band) [45], ...
The advent of large scale multimedia databases has led to great challenges in content-based image retrieval (CBIR). Even though CBIR is considered an emerging field of research, however it constitutes a strong background for new methodologies and systems implementations. Therefore, many research contributions are focusing on techniques enabling higher image retrieval accuracy while preserving low level of computational complexity. Image retrieval based on texture features is receiving special attention because of the omnipresence of this visual feature in most real-world images. This paper highlights the state-of-the-art and current progress relevant to texture-based image retrieval and spatial-frequency image representations. In particular, it gives an overview of statistical methodologies and techniques employed for texture feature extraction using most popular spatial-frequency image transforms, namely discrete wavelets, Gabor wavelets, dual-tree complex wavelet and contourlets. Indications are also given about used similarity measurement functions and most important achieved results. Comment: 19 pages, 11 figures, 2 tables
