Search

Camouflage plays an indispensable role in modern military. Generally, a camouflage pattern includes a plurality of shapes units, which are split by elemental colors. The fixed camouflage pattern is low adaptability and concealability for changeable battlefields. Moreover, the conventionally camouflage design is time and resource consuming for manual drawing. In this paper, we propose a dynamic camouflage synthesizing method. Firstly, the texture patterns of one class of battlefield images is extracted using convolutional transfer network. We use 3*3 convolution kernels to extract texture features. And the covariance matrix is used as loss function to calculate the loss of different image texture features. Colors construction and their distribution of the specified battlefield images is extracted according to a clustering-based algorithm. Finally, the statistical color units are embedded into the texture patterns. We assess the adaptability and the concealability of synthesized camouflage using Eye Tracking based criteria. Comparing to the People’s Liberation Army Type 87 woodland pattern and the digital camouflage pattern synthesized using our previous method, testing results of our synthesized camouflage patterns were better in different saccade indicators. This demonstrates camouflage pattern synthesize proposed got better concealment and the validity of proposed method. Besides, we also compared our method with local binary pattern for extracting texture features. The experiments results indicated the method proposed had better consistency to the scene images.

Current steganalyzers based on deep learning mostly adopt wider or deeper designs to improve detection performance. However, an overly complex network increases the training cost and is not conducive to its expansion and optimization. Moreover, steganalysis pays more attention to high-frequency information corresponding to the image texture. However, the deeper the network, the more likely it is to learn low-frequency information corresponding to the image content, which is inconsistent with the goal of steganalysis. In response to these problems, a multi-frequency residual deep convolutional neural network for steganalysis of color images called MFRNet is proposed in this paper. We apply the idea of multi-frequency analysis to steganography detection for the first time, effectively controlling the network scale. By designing columns of different depths, it can learn different frequency components of steganographic noise at the same time. The detection performance is better than the existing networks that only learn a single frequency component of steganographic noise at the same depth. Therefore, it can achieve a good detection performance with a lighter architecture. In addition, by designing residual basic blocks with different residual shortcuts, different scales of steganographic noise residuals can be calculated at the same time, which can effectively suppress the interference of image content, so as to better reduce the impact of steganography algorithm mismatch and payload mismatch than the existing methods. The experimental results on PPG-LIRMM-COLOR showed that the proposed MFRNet outperformed the state-of-the-art model WISERNet.

The color and contrast of objects in the image will be affected by meteorological factors, especially rain and snow will block part of the image, which will change the information contained in the image. Image restoration under bad weather conditions has practical application value. At present, most of the research focuses on the removal of fog, and the research on complex rain and snow is relatively less. Rain has more prominent features in gradient domain, and it is more distinct from non-rain image texture. In this paper, Generative Adversarial Networks is used to combine the information of image in gradient domain and spatial domain to get better performance of rain removal. Gradient aided coding is used in the generator to generate depth features that are more conducive to rain removal. In the discriminator, the gradient is used as an additional input to provide more recognizable rain and non-rain information, which enhances the discriminator’s ability to distinguish the image generated by the generator and the ground truth. By modifying the network structure of the expanded spatial pyramid pooling (ASPP), the abnormal rain removal results produced by the generator are reduced. Experimental results show that the proposed method improves the performance of rain removal and the visual quality of the generated image.

The positioning accuracy of the gravity-aided navigation is closely related to the selection of local gravity maps. In this paper, the local gravity maps are converted into 8-bit gray images. The navigability features of the local gravity maps are extracted by using the image texture feature analysis methods, such as the gray histogram complexity, the Sobel operator, and the gray level co-occurrence matrix, and the navigability comprehensive evaluation of each local gravity map is obtained by using the projection pursuit model. According to the nebula model, a gravitation field algorithm is proposed. The gravitation field algorithm, genetic algorithm, and firefly algorithm are used to obtain the optimal projection direction of the projection pursuit model, respectively. We compared the optimizing results and gave the navigability evaluation of local gravity maps that provide the basis for the selection of local gravity map. The comparison results show that the gravitation field algorithm has the best performance in obtaining the optimal projection direction, and the contributions of the navigability features in the navigability evaluation are most equal. Under the same experimental conditions, the local gravity map selected by the method proposed in this paper has the highest positioning accuracy and the best matching track.

Facade structural features can represent the overall framework of buildings. However, structural features extracted by the current methods contain a quantity of trivial unstructured information. In this study, we proposed an accurate extraction method for structural features of building facades through texture fusion. By performing texture fusion on building facade images, the interference of textural elements on structural feature extraction could be eliminated. After texture fusion, the line segment detector (LSD) algorithm is used to extract the structural features from the building facade images, and random sample consensus (RANSAC) is used to improve the continuity of structural features. The accuracy and effectiveness of the proposed method was demonstrated by comparing results with the state-of-the-art methods, such as LSD, MLSD, CannyLines, and MCMLSD. Value setting of three important parameters is discussed in detail. The imagery facade features extracted through the proposed method provide valuable support for many fields, such as image feature registration and 3D reconstruction of building surfaces.

In this paper, a new feature descriptor, named correlated primary visual texton histogram features (CPV-THF), for image retrieval is proposed. CPV-THF integrates the visual content and semantic information of the image by finding correlations among the colour, texture orientation, intensity, and local spatial structure information of an image. Based on texton theory, box-shaped structural elements are designed for image texture analysis. The colour, texture orientation, and intensity feature histograms that are proposed in CPV-THF are represented by correlated attributes of the co-occurrence matrix. The performance of the proposed descriptor is compared with those of state-of-the-art texture, colour, shape, and local-pattern-based CBIR descriptors. Experiments are conducted on multiple standard natural image datasets, namely, corel1k, corel5k, and corel10k. Experimental results indicate that the proposed descriptor outperforms various state-of-the-art descriptors, such as GLCM, MTH, MSD, MS-LSP, SED, STH, and MTSD.

Rain streaks in the air appear in various blurring degrees and resolutions due to different distances from their positions to the camera. Similar rain patterns are visible in a rain image as well as its multi-scale (or multi-resolution) versions, which makes it possible to exploit such complementary information for rain streak representation. In this work, we explore the multi-scale collaborative representation for rain streaks from the perspective of input image scales and hierarchical deep features in a unified framework, termed multi-scale progressive fusion network (MSPFN) for single image rain streak removal. For similar rain streaks at different positions, we employ recurrent calculation to capture the global texture, thus allowing to explore the complementary and redundant information at the spatial dimension to characterize target rain streaks. Besides, we construct multi-scale pyramid structure, and further introduce the attention mechanism to guide the fine fusion of this correlated information from different scales. This multi-scale progressive fusion strategy not only promotes the cooperative representation, but also boosts the end-to-end training. Our proposed method is extensively evaluated on several benchmark datasets and achieves state-of-the-art results. Moreover, we conduct experiments on joint deraining, detection, and segmentation tasks, and inspire a new research direction of vision task-driven image deraining. The source code is available at \url{https://github.com/kuihua/MSPFN}., CVPR 2020

Interactive fashion image manipulation, which enables users to edit images with sketches and color strokes, is an interesting research problem with great application value. Existing works often treat it as a general inpainting task and do not fully leverage the semantic structural information in fashion images. Moreover, they directly utilize conventional convolution and normalization layers to restore the incomplete image, which tends to wash away the sketch and color information. In this paper, we propose a novel Fashion Editing Generative Adversarial Network (FE-GAN), which is capable of manipulating fashion images by free-form sketches and sparse color strokes. FE-GAN consists of two modules: 1) a free-form parsing network that learns to control the human parsing generation by manipulating sketch and color; 2) a parsing-aware inpainting network that renders detailed textures with semantic guidance from the human parsing map. A new attention normalization layer is further applied at multiple scales in the decoder of the inpainting network to enhance the quality of the synthesized image. Extensive experiments on high-resolution fashion image datasets demonstrate that the proposed FE-GAN significantly outperforms the state-of-the-art methods on fashion image manipulation.

The ability to visually re-identify objects is a fundamental capability in vision systems. Oftentimes, it relies on collections of visual signatures based on descriptors, such as SIFT or SURF. However, these traditional descriptors were designed for a certain domain of surface appearances and geometries (limited relief). Consequently, highly-textured surfaces such as tree bark pose a challenge to them. In turn, this makes it more difficult to use trees as identifiable landmarks for navigational purposes (robotics) or to track felled lumber along a supply chain (logistics). We thus propose to use data-driven descriptors trained on bark images for tree surface re-identification. To this effect, we collected a large dataset containing 2,400 bark images with strong illumination changes, annotated by surface and with the ability to pixel-align them. We used this dataset to sample from more than 2 million 64x64 pixel patches to train our novel local descriptors DeepBark and SqueezeBark. Our DeepBark method has shown a clear advantage against the hand-crafted descriptors SIFT and SURF. For instance, we demonstrated that DeepBark can reach a mAP of 87.2% when retrieving 11 relevant bark images, i.e. corresponding to the same physical surface, to a bark query against 7,900 images. Our work thus suggests that re-identifying tree surfaces in a challenging illuminations context is possible. We also make public our dataset, which can be used to benchmark surface re-identification techniques.

Explosion phenomena today are considered a significant concern that needs to be detected and analyzed with a prompt response. We develop a multiclass categorization system for explosion phenomena using color images. Consequently, we describe four patterns of explosion phenomena, including pyroclastic density currents, lava fountains, lava and tephra fallout, and nuclear explosions, against three patterns of non-explosion phenomena, including wildfires, fireworks, and sky clouds. The classification task was handled through extracting different types of features, including texture features, amplitude features, frequency features, and histogram features. Then, these features were fed into several multiclass classification methods. In addition, we present a new data set for volcanic and nuclear explosions that includes 10 654 samples. Evaluation results show the one-against-one multiclass support vector machine with degree 3 polynomial kernel outperforms other classification methods. It produces the highest classification rate of 90.85% to categorize 5327 images of the data set. A reasonable execution time of approximately 117 ms was accomplished to classify one input test image.

Inspired by recent work on neural network image generation which rely on backpropagation towards the network inputs, we present a proof-of-concept system for speech texture synthesis and voice conversion based on two mechanisms: approximate inversion of the representation learned by a speech recognition neural network, and on matching statistics of neuron activations between different source and target utterances. Similar to image texture synthesis and neural style transfer, the system works by optimizing a cost function with respect to the input waveform samples. To this end we use a differentiable mel-filterbank feature extraction pipeline and train a convolutional CTC speech recognition network. Our system is able to extract speaker characteristics from very limited amounts of target speaker data, as little as a few seconds, and can be used to generate realistic speech babble or reconstruct an utterance in a different voice., Accepted to ICASSP 2018

Image dehazing is a technique to enhance the images acquired in poor weather conditions, such as fog and haze. Existing image dehazing methods are mainly based on dark channel prior. Since the dark channel is not reasonable for sky regions, a sky segmentation and region wised medium transmission based image dehazing method is proposed in this paper. Firstly, sky regions are segmented by quad-tree splitting based feature pixels detection. Then, a medium transmission estimation method for sky regions is proposed based on color characteristic observation of sky regions. The medium transmission is then filtered by an edge preserving guided filter. Finally, based on the estimated medium transmission, the hazed images are restored. Experimental results demonstrate that the performance of the proposed method is better than that of existing methods. The restored image is more natural, especially in the sky regions.

International audience; At the core of most three-dimensional alignment and tracking tasks resides the critical problem of point correspondence. In this context, the design of descriptors that efficiently and uniquely identifies keypoints, to be matched, is of central importance. Numerous descriptors have been developed for dealing with affine/perspective warps, but few can also handle non-rigid deformations. In this paper, we introduce a novel binary RGB-D descriptor invariant to iso-metric deformations. Our method uses geodesic isocurves on smooth textured manifolds. It combines appearance and geometric information from RGB-D images to tackle non-rigid transformations. We used our descriptor to track multiple textured depth maps and demonstrate that it produces reliable feature descriptors even in the presence of strong non-rigid deformations and depth noise. The experiments show that our descriptor outperforms different state-of-the-art descriptors in both precision-recall and recognition rate metrics. We also provide to the community a new dataset composed of annotated RGB-D images of different objects (shirts, cloths, paintings, bags), subjected to strong non-rigid deformations, to evaluate point correspondence algorithms .

Numerous environmental and financial benefits of using hyperspectral imaging have driven much increased applications on plant monitoring and diagnosis. This paper is concerned with analysis of hyperspectral images for plant discrimination by means of their spectral and texture properties. The main contribution of the work lies in the use feature selection and Markov random field model (MRF) to facilitate such spectral-texture analysis to enhance prediction performance, as compared to conventional analysis methods. A hyperspectral dataset on control and stressed Arabidopsis plant leaves captured by a proximal hyperspectral imaging system was used in the experiment. Texture parameters with different orders were estimated from the MRF model and two support vector machine settings were used in the evaluation. Experimental results showed significant improvements in classification performance of the proposed spectral-texture approach over the conventional analysis methods.

Hair is one of the elements that mostly characterize people appearance. Being able to detect hair in images can be useful in many applications, such as face recognition, gender classification, and video surveillance. To this purpose we propose a novel multi-class image database for hair detection in the wild, called Figaro. We tackle the problem of hair detection without relying on a-priori information related to head shape and location. Without using any human-body part classifier, we first classify image patches into hair vs. non-hair by relying on Histogram of Gradients (HOG) and Linear Ternary Pattern (LTP) texture features in a random forest scheme. Then we obtain results at pixel level by refining classified patches by a graph-based multiple segmentation method. Achieved segmentation accuracy (85%) is comparable to state-of-the-art on less challenging databases.

International audience; In this paper, we propose a combined feature approach which takes full advantages of local structure information and the more global one for improving texture image classification results. In this way, Local Binary Pattern is used for extracting local features, whilst the Scattering Transform feature plays the role of a global descriptor. Intensive experiments conducted on many texture benchmarks such as ALOT, CUReT, KTH-TIPS2-a, KTH-TIPS2b, and OUTEX show that the combined method outweigh each one which stands alone in term of classification accuracy. Also, our method outperforms many others, whilst it is comparable to state of the art on the experimented datasets.

Texture classification has been extensively studied in computer vision. Recent research shows that the combination of Fisher vector (FV) encoding and convolutional neural network (CNN) provides significant improvement in texture classification over the previous feature representation methods. However, by truncating the CNN model at the last convolutional layer, the CNN-based FV descriptors would not incorporate the full capability of neural networks in feature learning. In this study, we propose that we can further transform the CNN-based FV descriptors in a neural network model to obtain more discriminative feature representations. In particular, we design a locally-transferred Fisher vector (LFV) method, which involves a multi-layer neural network model containing locally connected layers to transform the input FV descriptors with filters of locally shared weights. The network is optimized based on the hinge loss of classification, and transferred FV descriptors are then used for image classification. Our results on three challenging texture image datasets show improved performance over the state-of-the-art approaches.

Reconstructing the detailed geometric structure of a face from a given image is a key to many computer vision and graphics applications, such as motion capture and reenactment. The reconstruction task is challenging as human faces vary extensively when considering expressions, poses, textures, and intrinsic geometries. While many approaches tackle this complexity by using additional data to reconstruct the face of a single subject, extracting facial surface from a single image remains a difficult problem. As a result, single-image based methods can usually provide only a rough estimate of the facial geometry. In contrast, we propose to leverage the power of convolutional neural networks to produce a highly detailed face reconstruction from a single image. For this purpose, we introduce an end-to-end CNN framework which derives the shape in a coarse-to-fine fashion. The proposed architecture is composed of two main blocks, a network that recovers the coarse facial geometry (CoarseNet), followed by a CNN that refines the facial features of that geometry (FineNet). The proposed networks are connected by a novel layer which renders a depth image given a mesh in 3D. Unlike object recognition and detection problems, there are no suitable datasets for training CNNs to perform face geometry reconstruction. Therefore, our training regime begins with a supervised phase, based on synthetic images, followed by an unsupervised phase that uses only unconstrained facial images. The accuracy and robustness of the proposed model is demonstrated by both qualitative and quantitative evaluation tests., Comment: 15 pages, supplementary material included

© 2017 IEEE. We present a method for synthesizing a frontal, neutralexpression image of a person's face given an input face photograph. This is achieved by learning to generate facial landmarks and textures from features extracted from a facial-recognition network. Unlike previous generative approaches, our encoding feature vector is largely invariant to lighting, pose, and facial expression. Exploiting this invariance, we train our decoder network using only frontal, neutral-expression photographs. Since these photographs are well aligned, we can decompose them into a sparse set of landmark points and aligned texture maps. The decoder then predicts landmarks and textures independently and combines them using a differentiable image warping operation. The resulting images can be used for a number of applications, such as analyzing facial attributes, exposure and white balance adjustment, or creating a 3-D avatar.

We propose a Deep Texture Encoding Network (Deep-TEN) with a novel Encoding Layer integrated on top of convolutional layers, which ports the entire dictionary learning and encoding pipeline into a single model. Current methods build from distinct components, using standard encoders with separate off-the-shelf features such as SIFT descriptors or pre-trained CNN features for material recognition. Our new approach provides an end-to-end learning framework, where the inherent visual vocabularies are learned directly from the loss function. The features, dictionaries and the encoding representation for the classifier are all learned simultaneously. The representation is orderless and therefore is particularly useful for material and texture recognition. The Encoding Layer generalizes robust residual encoders such as VLAD and Fisher Vectors, and has the property of discarding domain specific information which makes the learned convolutional features easier to transfer. Additionally, joint training using multiple datasets of varied sizes and class labels is supported resulting in increased recognition performance. The experimental results show superior performance as compared to state-of-the-art methods using gold-standard databases such as MINC-2500, Flickr Material Database, KTH-TIPS-2b, and two recent databases 4D-Light-Field-Material and GTOS. The source code for the complete system are publicly available.

Augmented Reality (AR) marker hiding is a technique to visually remove AR markers from a real-time video stream. The conventional approach transforms a background image with a homography matrix calculated on the basis of a camera pose and overlays the transformed image on the AR marker region in the real-time frame, assuming that the AR marker is on a planar surface. However, this approach can cause discontinuities in textures around the boundary between the marker and its surrounding area when the planar surface assumption is not satisfied. This paper proposes a method for AR marker hiding without discontinuities around texture boundaries under unknown and nonplanar background geometry. The proposed method estimates dense motion in the marker's background based on feature point correspondences around the marker, together with a smooth motion assumption, and deforms the background image according to the dense motion. Experimental results demonstrate the effectiveness of the proposed method in various environments with different background geometries and textures., ISMARW 2015 : IEEE International Symposium on Mixed and Augmented Reality Workshops , Sep 29-Oct 3, 2015 , Fukuoka, Japan

An approach to automatically detect and classify young spruce and birch trees in forest environment is presented. The method could be used in autonomous or semi-autonomous forest machines during tending operations. Detection is done by segmenting laser range images formed by a rotating laser scanner. Classification is done with a two-class Naive Bayes classifier based on image texture features. Multiple combinations of 99 features were tested and the best classifier included eight features from the co-occurrence matrix, local binary patterns, statistical geometrical features and Gabor filter. 79% of spruces and birches in the testing material were detected and 74% of these were correctly classified. Results suggest that the approach is suitable but there are still some challenges in each of the processing steps. Iteration between segmentation and classification is needed to increase reliability.

Previous work has shown that perceptual texture similarity and relative attributes cannot be well described by computational features. In this paper, we propose to predict human’s visual perception of texture images by learning a nonlinearmapping from computational feature space to perceptual space. Hand-crafted features and deep features, which were successfully applied in texture classification tasks, were extracted and used to train Random Forest and rankSVM models against perceptual data from psychophysical experiments. Three texture datasets were used to test our proposed method and the experiments show that the predictions of such learnt models are in high correlation with human’s results.

Image segmentation refers to the process of dividing an image into multiple regions which represent meaningful areas. Image segmentation is an essential step for most image analysis tasks such as object recognition and tracking, pattern recognition, content-based image retrieval, etc. In recent years, a large number of image segmentation algorithms have been developed, but achieving accurate segmentation still remains a challenging task. Recently, reservoir computing (RC) has drawn much attention in machine learning as a new model of recurrent neural networks (RNN). Echo State Network (ESN) represents one efficient realization of RC, which is initially designed to facilitate learning in Recurrent Neural Networks. In this paper we investigate the viability of ESN as feature extractor for pixel classification based colour image segmentation. Extensive experiments are conducted on real world colour image datasets and the global ESN reservoir parameters are varied to identify their operating ranges that allow the use of the reservoir nodes internal activations as new pixel features for the colour image segmentation task. A simple feed forward neural network is used to realize the ESN readout function and classify these new features. The experimental results show that the proposed method achieves high performance image segmentation comparing with state-of-the-art techniques. In addition, a set of empirically derived guidelines for setting the reservoir global parameters are proposed.

Rendering the semantic content of an image in different styles is a difficult image processing task. Arguably, a major limiting factor for previous approaches has been the lack of image representations that explicitly represent semantic information and, thus, allow to separate image content from style. Here we use image representations derived from Convolutional Neural Networks optimised for object recognition, which make high level image information explicit. We introduce A Neural Algorithm of Artistic Style that can separate and recombine the image content and style of natural images. The algorithm allows us to produce new images of high perceptual quality that combine the content of an arbitrary photograph with the appearance of numerous well-known artworks. Our results provide new insights into the deep image representations learned by Convolutional Neural Networks and demonstrate their potential for high level image synthesis and manipulation.

Content-aware image retrieval is a very important topic nowadays, when the amount of digital image data is highly increasing. Existing sketch based image retrieval (SBIR) systems perform at a reduced level on real life images, where background data may distort image descriptors and retrieval results. To avoid this, a preprocessing step is introduced in this paper to distinguish between foreground and background, using integrated saliency detection. To build the descriptor only on the most relevant pixels, orientation feature is extracted at salient Modified Harris for Edges and Corners (MHEC) keypoints using an improved edge map, resulting in a Salient Orientation Histogram (SOH). The proposed SBIR system is also augmented with a segmentation step for object detection. The method is tested on the THUR15000 database, containing random internet images. Image retrieval and object detection both give promising results compared to other state-of-the-art methods.

Image segmentation is still considered a very challenging subject despite years of research effort poured into the field. The problem is exacerbated when there is need for specific object detection. Since objects can be visually non-homogeneous, techniques that attempt to segment images into visually uniform regions using only the bottom-up cues, tend to fail. We propose a novel two-step model that incorporates both bottom-up information and top-down object constraints. Firstly, a set of uniform regions are generated using an extension of contour detection, seeded region growing, and graph-based methods. The second step applies co-occurrence constraints on the image regions in order to perceptually group regions into objects. This unsupervised segmentation process models each object using higher-level knowledge in the form of visual co-occurrences of its constituent parts. Experiments on the horse and ImageCLEF databases show that the proposed technique performs comparably well with existing state-of-the-art techniques.

We propose a method for construction of a cascaded traffic sign detector. Viola et al. have proposed a robust and extremely rapid object detection method based on a boosted cascade of simple feature classifiers. To obtain a high detection accuracy in real environment, it is necessary to train the classifier with a set of learning images which contain various appearances of detection targets. However, collecting the traffic sign images manually for training takes much cost. Therefore, we use a generative learning method for constructing the traffic sign detector. In this paper, shape, texture and color changes are considered in the generative learning. By this method, the performance of the traffic sign detection improves and the cost of collecting the training images is reduced at the same time. Experimental results using car-mounted camera images showed the effectiveness of the proposed method.

We investigate the influence of low-level image features for aesthetics prediction. We show that the aesthetic quality of a photography depends on its context. Image features learned from a specific image category are not necessarily the same as features learned from a generic image collection. Experiments conducted on specific image categories show that specific features obtain statistically significantly better results than generic ones.

Convolutional Neural Network (CNN) features have been successfully employed in recent works as an image descriptor for various vision tasks. But the inability of the deep CNN features to exhibit invariance to geometric transformations and object compositions poses a great challenge for image search. In this work, we demonstrate the effectiveness of the objectness prior over the deep CNN features of image regions for obtaining an invariant image representation. The proposed approach represents the image as a vector of pooled CNN features describing the underlying objects. This representation provides robustness to spatial layout of the objects in the scene and achieves invariance to general geometric transformations, such as translation, rotation and scaling. The proposed approach also leads to a compact representation of the scene, making each image occupy a smaller memory footprint. Experiments show that the proposed representation achieves state of the art retrieval results on a set of challenging benchmark image datasets, while maintaining a compact representation.

The objective of this paper is to design an embedding method mapping local features describing image (e.g. SIFT) to a higher dimensional representation used for image retrieval problem.

The orientation field of a fingerprint is crucial for feature extraction and matching. However, estimation of orientation fields in latents is very challenging because latents are usually of poor quality. Inspired by the superiority of convolutional neural networks (ConvNets) for various classification and recognition tasks, we pose latent orientation field estimation in a latent patch to a classification problem, and propose a ConvNet based approach for latent orientation field estimation. The underlying idea is to identify the orientation field of a latent patch as one of a set of representative orientation patterns. To achieve this, 128 representative orientation patterns are learnt from a large number of orientation fields. For each orientation pattern, 10,000 fingerprint patches are selected to train the ConvNet. To simulate the quality of latents, texture noise is added to the training patches. Given image patches extracted from a latent, their orientation patterns are predicted by the trained ConvNet and quilted together to estimate the orientation field of the whole latent. Experimental results on NIST SD27 latent database demonstrate that the proposed algorithm outperforms the state-of-the-art orientation field estimation algorithms and can boost the identification performance of a state-of-the-art latent matcher by score fusion.

This paper proposes an image processing technique based on based on Firefly Algorithm and simplified sobel filter in order to assist recognition procedure for 2D images. Experiments have been performed with the proposed solution applied on an open set of test images. Research results are presented and discussed and show the efficiency of implemented solution.

In this paper, we present our ongoing work on glaucoma classification using fundus images. The approach makes use of texture analysis based on Binary Robust Independent Elementary Features (BRIEF). This texture measurement is chosen because it can address the illumination issues of the retinal images and has a lower degree of computational complexity than most of the existing texture measurement methods currently used in the literature. Contrary to other approaches, the texture measures are extracted from the whole retina image without targeting any specific region. The method was tested on a set of 196 images composed of 110 healthy retina images and 86 glaucomatous images and achieved an area under curve (AUC) of 84%. A comparison performance with other texture measurements is also included, which shows our method to be superior.

High-resolution synthetic aperture sonar (SAS) systems yield richly detailed images of seabed environments. Algorithms that automatically segment and label seabed textures such as coral, sea grass, sand ripple, and mud, require suitable features that discriminate between the texture classes. Here we present a robust, parameterized SAS image texture model based on the autocorrelation function (ACF) of the intensity image. This ACF texture model has been shown to accurately model first- and second-order statistical features of various seabed environments. An unsupervised multi-class k-means segmentation algorithm that uses the features derived from the ACF model is employed to label rock and ripple textures from a set of textured SAS images. The results of the segmentation are compared against the performance of the segmentation approach using biorthogonal wavelets and Haralick features. In the described experiments, the ACF model features are shown to produce better segmentations than the features based on wavelet coefficients and Haralick features for classifiers of low complexity.

Structural information available from the granulometry of an image has been used widely in image texture analysis and classification. In this paper we present a method for classifying texture images which follow an intrinsic ordering of textures, using polynomial regression to express granulometric moments as a function of class label. Separate models are built for each individual moment and combined for back-prediction of the class label of a new image. The methodology was developed on synthetic images of evolving textures and tested using real images of 8 different grades of cut-tear-curl black tea leaves. For comparison, grey level co-occurrence (GLCM) based features were also computed, and both feature types were used in a range of classifiers including the regression approach. Experimental results demonstrate the superiority of the granulometric moments over GLCM-based features for classifying these tea images.

Due to the semantic gap, the automatic interpretation of digital images is a very challenging task. Both the segmentation and classification are intricate because of the high variation of the data. Therefore, the application of appropriate features is of utter importance. This paper presents biologically inspired texture features for material classification and interpreting outdoor scenery images. Experiments show that the presented texture features obtain the best classification results for material recognition compared to other well-known texture features, with an average classification rate of 93.0%. For scene analysis, both a bottom-up and top-down strategy are employed to bridge the semantic gap. At first, images are segmented into regions based on the perceptual texture and next, a semantic label is calculated for these regions. Since this emerging interpretation is still error prone, domain knowledge is ingested to achieve a more accurate description of the depicted scene. By applying both strategies, 91.9% of the pixels from outdoor scenery images obtained a correct label.

There are a wide variety of measures in the literature that capture the “coarseness” texture property. Some of them have better ability to represent coarseness than the others. Furthermore, some of them are more robust against the variation of other image features, like brightness, contrast, noise and size of the image. In this paper, we propose to study the robustness and the relationship with human coarseness perception of 17 classical measures of coarseness, in order to obtain a ranking of measures. This ranking can be used to identify those measures that have the highest relationship degree with perception and the least variation with the other image features.

Image processing plays an important role in extracting useful information from images. However, the image processing and the process of translating an image into a statistical distribution of low-level features is not an easy task. These tasks are complicated since the acquired image data often noisy, and target objects are influenced by lighting, intensity or illumination. In the case of flower classification, image processing is a crucial step for computer-aided plant species identification. Flower image classification is based on the low-level features such as colour and texture to define and describe the image content. Colour features are extracted using normalized colour histogram and texture features are extracted using gray-level co-occurrence matrix. In this study, a dataset consists of 180 patterns with 7 attributes for each type of flower has been gathered. The finding from the study reveals that the number of images generated to represent each type of flower influences the classification accuracy. One interesting observation is that duplication of very hard to learn images assist Neural Network to improve its classification accuracy. This is also another area that could lead to better understanding towards the behaviour of images when applied to Neural Network classification.

© 2014 IEEE. In this paper, we focus on developing a novel noise-robust LBP-based texture feature extraction scheme for texture classification. Specifically, two solutions have been proposed to overcome the primary two reasons that cause local binary pattern sensitive to noise. First, a hybrid model is proposed for noise-robust texture description. In this new model, the local primitive micro features are encoded with the texture's global spatial structure to reduce the noise sensitiveness. Second, we design an adaptive quantization algorithm, in which quantization thresholds are choosing adaptively on the basis of the texture's content. Higher noise-tolerance and discriminant power can be obtained in the quantization process. Based on the proposed hybrid texture description model and adaptive quantization algorithm, we develop an adaptive hybrid pattern scheme for noise-robust texture feature extraction. Compared with several state-of-the-art feature extraction schemes, our scheme leads to significant improvement in noisy texture classification.

Since feature extraction not only plays a key role, but also have a great effect on image analysis and registration. In this paper, we evaluate a comparative performance of three point feature extraction algorithm such as SIFT, SURF and BRISK. The experimental results show that SURF exhibited a better significant results on extracting point features and matching time, compared to SIFT and BRISK. The objective of this research is to provide a decision support which one should be chosen and acceptable among point-based image descriptors in the industrial field.

We propose a new approach for constructing mid-level visual features for image classification. We represent an image using the outputs of a collection of binary classifiers. These binary classifiers are trained to differentiate pairs of object classes in an object hierarchy. Our feature representation implicitly captures the hierarchical structure in object classes. We show that our proposed approach outperforms other baseline methods in image classification.

In this paper, we deal with the problem of categorizing different underwater habitat types. Previous works on solving this categorization problem are mostly based on the analysis of underwater images. In our work, we design a system capable of categorizing underwater habitats based on underwater video content analysis since the temporally correlated information may make contribution to the categorization task. However, the task is very challenging since the underwater scene in the video is continuously varying because of the changing scene and surface conditions, lighting, and the viewpoint. To that end, we investigate the utility of two approaches to underwater video classification: the common spatio-temporal interest points (STIPs) and the video texture dynamic systems, where we model the underwater footage using dynamic textures and construct a categorization framework using the approach of the Bag-of-Systems(BoSs). We also introduce a new underwater video data set, which is composed of more than 100 hours of annotated video sequences. Our results indicate that, for the underwater habitat identification, the dynamic texture approach has multiple benefits over the traditional STIP-based video modeling.

Object detection in images withstanding significant clutter and occlusion is still a challenging task whenever the object surface is characterized by poor informative content. We propose to tackle this problem by a compact and distinctive representation of groups of neighboring line segments aggregated over limited spatial supports and invariant to rotation, translation and scale changes. Peculiarly, our proposal allows for leveraging on the inherent strengths of descriptor-based approaches, i.e. robustness to occlusion and clutter and scalability with respect to the size of the model library, also when dealing with scarcely textured objects.

In this paper we propose a novel framework for object retrieval based on automatic foreground object extraction and multi-layer information integration. Specifically, user interested objects are firstly detected from unconstrained videos via a multimodal cues method, then an automatic object extraction algorithm based on Grab Cut is applied to separate foreground object from background. The object-level information is enhanced during the feature extraction layer by assigning different weights to foreground and background pixels respectively, and the spatial color and texture information is integrated during the similarity calculation layer. Experimental results on both benchmark data set and real-world data set demonstrate the effectiveness of the proposed framework.

Attributes are helpful to infer high-level semantic knowledge of pedestrians, thus improving the performance of pedestrian tracking, retrieval, re-identification, etc. However, current pedestrian databases are mainly for the pedestrian detection or tracking application, and semantic attribute annotations related to pedestrians are rarely provided. In this paper, we construct an Attributed Pedestrians in Surveillance (APiS) database with various scenes. The APiS 1.0 database includes 3661 images with 11 binary and 2 multi-class attribute annotations. Moreover, we develop an evaluation protocol for researchers to evaluate pedestrian attribute classification algorithms. With the APiS 1.0 database, we present two baseline methods, one for binary attribute classification and the other for multi-class attribute classification. For binary attribute classification, we train AdaBoost classifiers with color and texture features, while for multi-class attribute classification, we adopt a weighted K Nearest Neighbors (KNN) classifier with color features. Finally, we report and discuss the baseline performance on the APiS 1.0 database following the proposed evaluation protocol.