Search

We propose an image upsampling algorithm to increase the depth image resolution while preserving edge information and protecting from the texture copy problem. The proposed upsampler selectively combines the spatial, depth range, and color range weight functions for cost calculation according to the weight selection algorithm. In particular, a color range weight function is proposed to reduce the texture copy problem caused by distinct patterns in the color image. Simulation results show that the proposed algorithm outperforms other conventional upsamplers.

Aiming at the problem of low segmentation accuracy and high computation time by applying existing segmentation methods for underwater color images, this paper has proposed an underwater color image segmentation method via RGB color channel fusion. Based on thresholding segmentation methods to conduct fast segmentation, the proposed method relies on dynamic estimation of the optimal weights for RGB channel fusion to obtain the grayscale image with high foreground-background contrast and reaches high segmentation accuracy. To verify the segmentation accuracy of the proposed method, the authors have conducted various underwater comparative experiments. The experimental results demonstrate that the proposed method is robust to illumination, and it is superior to existing methods in terms of both segmentation accuracy and computation time. Moreover, a segmentation technique is proposed for image sequences for real-time autonomous underwater vehicle operations.

In fringe projection profilometry, the measurement system generally consists of a projector for casting fringes onto a measured object and a monochrome camera for capturing the deformed fringe patterns. In addition to these components, we can add a color camera for capturing the texture of the object simultaneously. For implementing texture mapping on the reconstructed three-dimensional (3-D) surface, the parallax between the views of the texture camera and the measuring camera has to be corrected. For this purpose, we analyze the geometry of the fringe projection system with a color texture camera and further suggest a system calibration method. Using this method, the corresponding relationship between the texture and the 3-D data and the mapping relationship between the depths and the fringe phases are determined simultaneously, so that the duration time for the system calibration implementation is saved. The data processing with this method is of a low computational complexity because it involves only linear minimizations. Using the calibration results, we can transform the texture image from the view of the color camera to that of the measuring camera and precisely map it on the reconstructed object surface. Experimental results demonstrate that this method is effective in correcting the parallax of the texture image.

Infrared texture is an important feature in identifying scenery. To simulate infrared image texture effectively at different distances, we propose a model of infrared image texture generation based on scenery space frequency and the image pyramid degradation principle. First, we build a spatial frequency filter model based on imaging distance, taking into account the detector’s maximum spatial frequency, and use the filter to process a “zero” distance infrared image texture. Second, taking into consideration the actual temperature difference of the scenery’s details due to variation of the imaging distance and the effect of atmospheric transmission, we compare the actual temperature difference with the minimum resolvable temperature difference of the thermal imaging system at a specific frequency and produce a new image texture. The results show that the simulated multiresolution infrared image textures produced by the proposed model are very similar (lowest mean square error=0.51 and highest peak signal-to-noise ratio=117.59) to the images captured by the thermal imager. Therefore, the proposed model can effectively simulate infrared image textures at different distances.

We propose a novel multistage smoke detection algorithm based on inherent optical characteristics such as diffusion, color, and texture of smoke. Moving regions in a video frame are detected by an approximate median background subtraction method using the diffusion behavior of smoke. These moving regions are segmented by a fuzzy C-means (FCM) clustering algorithm that uses the hue and saturation components of moving pixels in the hue-saturation-intensity color space. A decision rule is used to select candidate smoke regions from smoke-colored FCM clusters. An object tracking approach is employed in the candidate smoke region to detect candidate smoke objects in the video frame, and image texture parameters are extracted from these objects using a gray level co-occurrence matrix (GLCM). The thirteen GLCM features are selected to constitute the feature vector by applying principal components analysis, resulting in high-accuracy smoke detection. Finally, a back propagation neural network is utilized as a classifier to discriminate smoke and nonsmoke using the selected feature vector. Experimental results using a standard experimental dataset of video clips demonstrate that the proposed approach outperforms state-of-the-art smoke detection approaches in terms of accuracy, making real-life implementation feasible.

A novel method to correct texture variations resulting from scale magnification, narrowing caused by cropping into the original size, or spatial rotation is discussed. The variations usually occur in images captured by a camera using different focal lengths. A representative region-matched algorithm is developed to improve texture classification after magnification, narrowing, and spatial rotation. By using a minimum ellipse, a representative region-matched algorithm encloses a specific region extracted by the J-image segmentation algorithm. After translating the coordinates, the equation of an ellipse in the rotated texture can be formulated as that of an ellipse in the original texture. The rotated invariant property of ellipse provides an efficient method to identify the rotated texture. Additionally, the scale-variant representative region can be classified by adopting scale-invariant parameters. Moreover, a hybrid texture filter is developed. In the hybrid texture filter, the scheme of texture feature extraction includes the Gabor wavelet and the representative region-matched algorithm. Support vector machines are introduced as the classifier. The proposed hybrid texture filter performs excellently with respect to classifying both the stochastic and structural textures. Furthermore, experimental results demonstrate that the proposed algorithm outperforms conventional design algorithms.

Over the past four decades, the satellite imaging sensors have acquired huge quantities of Earth- observation data. Content-based image retrieval allows for fast and effective queries of remote sensing images. Here, we take the following two issues into consideration. Firstly, different features and their combination should be chosen for different land covers. Secondly, for the block dividing strategy and the complexities of the remote sensing images, it can not effectively retrieve some small target areas scattered in multiple nontarget blocks. Aiming at the above two issues, a new region-based retrieval method with adaptive image segmentation is proposed. In order to improve the accuracy of remote sensing image segmentation, feature selection and weighing is performed by two-stage clustering, and image segmentation is accomplished based on the chosen features and mean shift procedure. Meanwhile, for the homogeneous characteristics of remote sensing land covers, a new regional representation and matching scheme are adopted to perform image retrieval. Experimental results on retrieving various land covers show that the method can avoid the impact of traditional blocking strategies, and can achieve an average percentage of 19% higher precision with the same level of recall rate, than the relevance feedback method for small target areas.

Spectral image complexity is an ill-defined term that has been addressed previously in terms of dimensionality, multivariate normality, and other approaches. Here, we apply the concept of the linear mixture model to the question of spectral image complexity at spatially local scales. Essentially, the "complexity" of an image region is related to the volume of a convex set enclosing the data in the spectral space. The volume is estimated as a function of increasing dimensionality (through the use of a set of endmembers describing the data cloud) using the Gram Matrix approach. It is hypothesized that more complex regions of the image are composed of multiple, diverse materials and will thus occupy a larger volume in the hyperspace. The ultimate application here is large area image search without a priori information regarding the target signature. Instead, image cues will be provided based on local, relative estimates of the image complexity. The technique used to estimate the spectral image complexity is described and results are shown for representative image chips and a large area flightline of reflective hyperspectral imagery. The extension to the problem of large area search will then be described and results are shown for a 4-band multispectral image.

Detecting the vanishing point from a single road image is a challenging problem because there is very limited information in the input image that can help the computer to deduce the genuine location of vanishing point. Besides, the cluttered ambient environment in a real road image sometimes will hinder rather than assist the detection. Learning both the advantages and the limitations of current edge-based and texture-based approaches motivates us to propose a new vanishing point detection method that exploits the intrinsic geometric line structures and color texture properties of general roads. Our approach integrates the efficiency of line segments of edge-based methods, and the orientation coherence concept that is frequently applied in texture-based methods, which can be of great help to improve the accuracy of selecting the right line segments for vanishing point detection. The proposed method has been implemented and tested on over 1000 various road images. These road images exhibit large variations in color, texture, illumination condition, and ambient environment. The experimental results demonstrate that this new method is both efficient and effective in detecting vanishing point when compared to the state-of-the-art edge-based and texture-based methods.

We propose an innovative and efficient approach to improve K-view-template (K-view-T) and K-view-datagram (K-view-D) algorithms for image texture classification. The proposed approach, called the weighted K-view-voting algorithm (K-view-V), uses a novel voting method for texture classification and an accelerating method based on the efficient summed square image (SSI) scheme as well as fast Fourier transform (FFT) to enable overall faster processing. Decision making, which assigns a pixel to a texture class, occurs by using our weighted voting method among the "promising" members in the neighborhood of a classified pixel. In other words, this neighborhood consists of all the views, and each view has a classified pixel in its territory. Experimental results on benchmark images, which are randomly taken from Brodatz Gallery and natural and medical images, show that this new classification algorithm gives higher classification accuracy than existing K-view algorithms. In particular, it improves the accurate classification of pixels near the texture boundary. In addition, the proposed acceleration method improves the processing speed of K-view-V as it requires much less computation time than other K-view algorithms. Compared with the results of earlier developed K-view algorithms and the gray level co-occurrence matrix (GLCM), the proposed algorithm is more robust, faster, and more accurate.

This paper presents a novel colony analysis system including an adjustable image acquisition subsystem and a wavelet-watershed-based image segmentation algorithm. An illumination box was constructed-both front lightning and back lightning illuminations can be chosen by users based on the properties of Petri dishes. In the illumination box, the lightning is uniform, which makes image processing easy. A digital camera at the top of the box is connected to a PC computer; all the camera functions are controlled by the developed computer software in this study. As usual, in the image processing part, the hardest task is image segmentation which is carried out by the four different algorithms: 1. recursive image segmentation on gray similarity; 2. canny edge detection-based segmentation; 3. the combination of 1 and 2, and 4. colony delineation on wavelet and watershed. The first three algorithms can obtain good results for ordinary colony images, and for the images including a lot of small (tiny) colonies and dark colonies and overlapping (or touching) colonies, the algorithm 4 can obtain better results than the others. The algorithms are tested by using a large number of different colony images, and the testing results are satisfactory.

Removing shadows from single color images is an important problem in computer vision. In this paper, we propose a novel shadow removal approach, which could effectively remove shadows from textured surfaces, yielding high quality shadow-free images. Our approach aims at calculating scale factors to cancel the effect of shadows. Based on the regional gray edge hypothesis, which assumes the average of the reflectance differences in a region is achromatic, the scale factors can be computed without the restrictions that former algorithms need. The experimental results show that the proposed algorithm is effective.

A structured light system simplifies three-dimensional reconstruction by illuminating a specially designed pattern to the target object, thereby generating a distinct texture on it for imaging and further processing. Success of the system hinges upon what features are to be coded in the projected pattern, extracted in the captured image, and matched between the projector's display panel and the camera's image plane. The codes have to be such that they are largely preserved in the image data upon illumination from the projector, reflection from the target object, and projective distortion in the imaging process. The features also need to be reliably extracted in the image domain. In this article, a two-dimensional pseudorandom pattern consisting of rhombic color elements is proposed, and the grid points between the pattern elements are chosen as the feature points. We describe how a type classification of the grid points plus the pseudorandomness of the projected pattern can equip each grid point with a unique label that is preserved in the captured image. We also present a grid point detector that extracts the grid points without the need of segmenting the pattern elements, and that localizes the grid points in subpixel accuracy. Extensive experiments are presented to illustrate that, with the proposed pattern feature definition and feature detector, more features points in higher accuracy can be reconstructed in comparison with the existing pseudorandomly encoded structured light systems.

Remote sensing provides a new idea and an advanced method for lithology identification, but lithology identification by remote sensing is quite difficult because 1. the disciplines of lithology identification in a concrete region are often quite different from the experts' experience; 2. In the regions with flourishing vegetation, lithology information is poor, so it is very difficult to identify the lithologies by remote sensing images...An intelligent method proposed in this paper for lithology identification based on support vector machine (SVM) and adaptive cellular automata (ACA) is expected to solve the above problems. The method adopted Landsat-7 ETM+ images and 1:50000 geological map as the data origins. It first derived the lithology identification factors on three aspects: 1. spectra, 2. texture and 3. Vegetation cover. Second, it plied the remote sensing images with the geological map and established the SVM to obtain the transition rules according to the factor values of the samples. Finally, it established an ACA model to intelligently identify the lithologies according to the transition and neighborhood rules. In this paper an ACA model is proposed and compared with the traditional one.

Denoising is expected to improve the image quality and the performance of analysis. Combining the nonlocal-means filter and the mean-shift method, we derive a nonlocal weights estimator considering not only similarity of intensity but also the spatial relationship between image blocks. Filtering is performed in each image block and the center pixel is restored by an oriented filter. Experimental results show the performance of the proposed method in improving the signal-to-noise ratio and preserving local textures.

Image denoising while preserving image features is a key problem in image processing and computer vision. This letter proposes an adaptive mixed method for image restoration. First, this method decomposes a given image as the sum of two components: geometric structure and oscillating pattern according to Meyer's theory. Second, a coupled bidirectional diffusion equation is used to restore the structure part, and a nonlocal means filter is used to remove noise in the oscillating part. Experimental results show advantages of this method in feature-preserving denoising.

We propose an approach for textured image segmentation based on amplitude-modulation frequency-modulation models. An image is modeled as a set of 2-D nonstationary sinusoids with spatially varying amplitudes and spatially varying frequency vectors. First, the demodulation procedure for the models furnishes a high-dimensional output at each pixel. Then, features including texture contrast, scale, and brightness are elaborately selected based on the high-dimensional output and the image itself. Next, a normalization and weighting scheme for feature combination is presented. Finally, simple K-means clustering is utilized for segmentation. The main characteristic of this work provides a feature vector that strengthens useful information and has fewer dimensionalities simultaneously. The proposed approach is compared with the dominant component analysis (DCA)+K-means algorithm and the DCA+ weighted curve evolution algorithm on three different datasets. The experimental results demonstrate that the proposed approach outperforms the others.

Watermark embedding processes usually result in some deg- radation of the digital multimedia contents. Hence, they are not suitable for valuable and sensitive digital multimedia contents. Therefore, schemes combining signature with digital watermarking-like techniques have been proposed. Based on the general model for these combined schemes, a new scheme for image copyright protection by utilizing the pinned field of the protected image is proposed in this paper. The pinned field explores the texture information of the images and can be used to enhance the watermark robustness. Experimental results show that the proposed scheme works well under different signal-processing and geo- metric transformation attacks. On the other hand, in comparison with related schemes in the literature, our proposed scheme also has better performance. Furthermore, the proposed scheme is robust to both JPEG lossy compression and additive Gaussian noise. © 2010 Society of Photo

We propose a fusion method for multiple patch-based depth image completion via local stereo warping, in order to refine the disparity map. The false disparities are often extracted from occluded areas and textureless areas in stereo matching due to their ambiguities. Our fusion method iteratively performs the following three steps in order to reconstruct the areas with false disparities. First, we generate a reliability map from the initial disparity map. Second, we extract multiple patch candidates by using the proposed depth image completion. Finally, we verify each patch candidate by using local stereo warping in order to update the reliability map. Experimental results show that the proposed method is very robust for large textureless areas and occluded areas as well because our method yields superior results in comparison to many state-of-the-art methods using stereo matching.

The criterion of an evaluation method for fused images is critical to the performance of different image fusion algorithms. We present novel metrics for evaluation of fused images, based on the similarity of corresponding regions in images. The new metrics are computed on a region-by-region basis that is more suitable for evaluation, because human eyes are more sensitive to regions. The region information is represented by a feature matrix of the region, which consists of multifeature vectors including spatial information, texture, and gray value, which can adequately reflect the regional content. These make evaluation methods from the pixel level to the feature level. Research indicates that the proposed metrics are more consistent with the nature of human perception, as it considers the local image variations and the saliency of region. Experimental results show the effectiveness of the proposed metrics for the evaluation of fused images.

A multiple color-filter aperture (MCA) can provide a single camera with depth information and multifocusing. However, the original version of the MCA system exhibits inherent limitations such as manual, empirical tuning parameters for the color channel registration and fusion (CRF) process. Furthermore, a CRF output image still contains undesired out-of-focus blur because of the finite-sized apertures and the lateral displacement of each color-filter aperture, which results in low exposure, color mixing, deviation of color convergence, and divergence of light rays. For overcoming these problems, we present a real-time image processing solution for digital multifocusing in a MCA system.

A measure for objectively assessing the performance of color image fusion methods is proposed. Two different aspects are considered in establishing the proposed measure-namely, the amount of common information between the source images and the final fused image as well as the distribution of color information in the final image in order to achieve optimal color representation. Mutual information and conditional mutual information are employed in order to assess information transfer between the source images and the final fused image. Simultaneously, the distribution of colors in the final image is explored by means of the hue coordinate in the perceptually uniform CIELAB space. The proposed measure does not depend on the use of a target-fused image for the objective performance evaluation. It is employed experimentally for objective evaluation of fusion methods in the cases of medical imaging and night vision data.

Live-wire-based image segmentation allows the user to select piece-wise globally optimal contour segments corresponding to a desired object boundary. However, the boundary extracted by live wire tends to deviate from the object boundary if other boundaries with strong gradient features are situated near a desirable object boundary. We present a new live-wire method based on the concept of embedded edge confidence to improve segmentation performance. Experimental results on medical images demonstrate the effectiveness of the proposed method.

Scalable video coding, which is an extension of H.264/AVC, provides efficient methods for merging 8-bits/pixel sequences and higher bits per pixel sequences into one scalable bitstream. In this work, two types of prediction approaches are supported. One is based on interlayer intraprediction, by which texture residuals between the base layer and the enhancement layer are removed. The interlayer intraprediction is restricted to the enhancement macroblock whose collocated macroblock is intracoded. In this prediction structure, a global look-up table, obtained from an arithmetic average or scale offset pairs of each macroblock, is chosen as an inverse tone-mapping process, depending on rate distortion (RD) performance. The other is based on interlayer motion prediction, where motion compensation is performed in the enhancement layer as well as in the base layer. But motion parameters in the enhancement layer are inferred from the corresponding blocks in the base layer, which minimizes the complexity increase. A coding gain of 0.14 dB or 1.2% reduction in bit rate is obtained for 10 bits/pixel test sequences, while a coding gain of 12 bits/pixel test sequences reaches up to 4.2 dB or 48% reduction in bit rate.

Rotation-invariant texture classification is one of the most challenging problems in computer vision. We present a new and effective method for rotation-invariant texture classification based on circular Gabor wavelets. Two group features can be constructed by the mean and variance of the circular Gabor filtered images and rotation invariants. Using the two group features, a discriminant can be found to classify rotated images. The proposed method is evaluated on three public texture databases: Brodatz, CUReT, and UIUCTex. The experimental results, based on different testing data sets, show that the proposed method has comparatively high correct classification rates not only for the rotated images, but also for the images under different illuminations and viewing directions. The proposed method is robust to additive white noise.

Extracting effective features for texture classification is always a difficult problem in texture analysis. We present a method for texture features extraction by independent component analysis (ICA) of Gabor features (called ICAG). It has three distinguished aspects. First, Gabor wavelet transformation first produces distinct textural features characterized by spatial locality, scale, and orientation selectivity. Second, principal component analysis (PCA) and ICA then reduce the dimensionality and redundancy of these features. Thus, the resulting independent components are taken as texture features for classification. Third, in the ICA procedure, two different frameworks are discussed. Framework I, called ICAG I, regards pixels as random variables and represents them as a column vector by reshaping all the transformed images row by row, while framework II, called ICAG II, treats the statistical features, namely, the mean and standard deviation, as random variables. Thus, the statistical features of all the transformed images construct a column vector. Finally, comparative experiments between ICAG and the other traditional methods—ICA and Gabor wavelets—are performed on two composite images (five narrowband textures and five natural textures with broader bands). The results indicate that ICAG provides the best performance and ICAG II is more efficient and applicable.

A system for simultaneous real-time three-dimensional geometry and color texture acquisition. The system includes a system processor for generating at least three phase shifted black and white fringe patterns with a phase shift of 2 π/3, a light projector adapted to project the fringe patterns onto an object, the projector being electrically connected with the system processor, and a color camera for capturing the fringe patterns to generate at least three raw fringe images. The fringe images are used to calculate a black and white texture image which is further converted to a color image by employing a demosaicing algorithm. The fringe images are also used to calculate a wrapped phase map that is further processed to generate a continuous unwrapped phase map by employing a phase unwrapping algorithm and the unwrapped phase map is converted to co-ordinates using calibrated system parameters for point-by-point three-dimensional shape measurement.

We perform unsupervised image classification based on texture features by using a novel evolutionary clustering method, named manifold evolutionary clustering (MEC). In MEC, the clustering problem is considered from a combinatorial optimization viewpoint. Each individual is a sequence of real integers representing the cluster representatives. Each datum is assigned to a cluster representative according to a novel manifold-distance-based dissimilarity measure, which measures the geodesic distance along the manifold. After extracting texture features from an image, MEC determines partitioning of the feature vectors using evolutionary search. We apply MEC to solve seven benchmark clustering problems on artificial data sets, three artificial texture image classification problems, and two synthetic aperture radar image classification problems. The experimental results show that in terms of cluster quality and robustness, MEC outperforms the K-means algorithm, a modified K-means algorithm using the manifold-distance-based dissimilarity measure, and a genetic-algorithm-based clustering technique in partitioning most of the test problems.

We propose an a posteriori kernel orthogonal subspace technique to segment texture images. It is a nonlinear version of the signature subspace classifier (SSC) derived on the basis of an unconstrained least-square estimation. In this approach, the linear subspace mixture model for the SSC is first reformulated in feature space via nonlinear mapping. Then the SSC in feature space is kernelized in terms of the kernel functions so that the dot products in the high dimensional feature space can be implicitly calculated by kernels. The obtained kernel SSC (KSSC) is equivalent to a nonlinear SSC in the input space. After that, the KSSC is applied to segment the texture images. To reduce the computational requirement in segmentation, the multiresolution-based technique (MKSSC) is developed. Experimental results demonstrate that the proposed MKSSC approach can effectively segment texture images and outperforms the MSSC method.

With the increasing needs in security systems, iris recognition is reliable as one of the important solutions for biometrics-based identification systems. This work presents an effective approach for iris recognition by analyzing iris patterns. To improve the rate of recognition, we divide the normalized iris image into several regions to keep the iris image away from several noise factors, such as eyelids, eyelashes, and motion blur. For feature extraction, the local edge pattern (LEP) operator is designed to capture local characteristics of the iris image to produce discriminating texture features in every region. A resulting 2D feature vector is mapped into a low-dimensional subspace using two dimension linear discriminant analysis (2DLDA), and then the minimum distance classifier (MDC) is adopted for recognition. Experiments on the public and freely available iris images taken from the CASIA (Institute of Automation, Chinese Academy of Sciences) and UBIRIS databases confirm the advantage of the proposed approach in terms of speed and accuracy.

The usual methods of generating image mosaics suffer from abrupt color change where the color varies little in the source image, and from discontinuity of texture or shape where they are continuous in the source image. Our proposed method is based on wavelet transforms. It not only considers the similarity between a single tile and the corresponding original grid image, but also takes into account the match of surrounding tiles. The results of our experiments show that color changes are gentler in areas with similar colors originally, shapes are smoother, and textures are more refined in our image mosaics.

This paper presents a novel multiresolution approach for texture classification based on the moment features, which are extracted in the x direction and y direction, of a histogram in each image resolution. Then we propose a weighted multiresolution moment feature for supervised texture classification. Moreover, because of their capability of expressing image information using moment methods, these features achieved better performance on rotation invariance and noise robustness than most of popular texture features presented in recent years.

We propose a novel deringing scheme based on the concept of a trilateral filter—an extension of the bilateral filter that consists of domain, range, and texture activity filters. The domain and range filters are used to measure image geometrical similarity and photometric similarity, respectively, while the texture activity filter helps to preserve image details during postprocessing. An image texture map is first computed with a dual-edge detection process. This texture map is then utilized to adjust the trilateral filter for effectively protecting the image detail from oversmoothing. Experimental results are provided to justify the algorithm.

We propose a novel automatic seeded region growing method based on gradient vector flow (GVF) for color image segmentation. YCbCr color space is selected to avoid the high correlation of RGB color space. First, a GVF field is constructed from an edge map of the input image. Then a scaler force field is derived from it by minimizing an energy functional iteratively. From the scalar field, we can select a set of seeds and get an initial segmentation via a straightforward downstream process. Finally, a region adjacency graph–based region merging is applied to merge similar neighboring regions into true results. Experimental results demonstrate that this method is insensitive to noises and efficient to multiple objects segmentation in color images.

We have developed a mapping algorithm for correcting sinusoidally scanned images from their distortions. Our algorithm is based on the close relationship between linear and sinusoidal scanning. Straightforward implementation of this algorithm showed that the mapped image has either missing lines or redundant lines. The missing lines were filled by fusing the mapped image with its median-filtered or interpolated version. The implementation of this algorithm shows that it is possible to retrieve up to 98% of the original image (depending on the algorithm used for data fusion) as measured by the recovered energy. Excellent correction was obtained for both simulated scanned images and actual images from a scanning laser radar system.

We present a novel texture feature persistence metric for automatic-target-recognition (ATR)-directed image compression based on the similarity between shapes. On the basis of spatial fuzzy representation of shapes, a similarity metric between shapes is proposed. Then the impact of lossy image compression on ATR performance is measured by the similarity between shapes, which are obtained by identical segmentation and edge extraction of the source image and degraded image after compression. Experimental results show that this metric effectively measures the extent to which target texture features are preserved after compression.

A new method for color image segmentation is proposed. It is based on a novel region-growing technique with a growth tolerance parameter that changes with step size, which depends on the variance of the actual grown region. Contrast is introduced to determine which value of the tolerance parameter is taken, choosing the one that provides the region with the highest contrast in relation to the background. Color and texture information are extracted from the image by means of a novel idea: the construction of a color distance image and a texture energy image. The color distance image is formed by calculating CIEDE2000 distance in the L*a*b* color space. The texture energy image is extracted from some statistical moments. Then, a novel texture-controlled multistep region-growing process is performed for the segmentation. One advantage of the method is that it is not designed to work with a particular kind of images. This method is tested on 80 natural color images of the Corel photo stock collection with excellent results. Numerical evidence of the quality of these results is provided by comparing them with the manual segmentation of five experts and with another color and texture segmentation algorithm.

A novel method of color and texture feature extraction for image retrieval is proposed. First, an input image is divided into 4×4 nonoverlapping blocks. Then, each block is quantized into two colors by color moment preserving and merging similar colors according to their color difference values under a preset threshold value. A few dominant colors are obtained, and a color histogram of the reconstructed image is computed for the color feature. For the texture feature, by preserving some color moments in the 4×4 neighborhood of each pixel, a bitmap is generated, which can be matched with 64 predefined bitmaps. Then a bitmap spectrum is obtained by counting bitmaps frequencies when all pixels of the input image are considered. An efficient method of similarity measurement is also presented. Experimental results show the proposed method performs well in terms of precision and recall.

A multimedia database system should deal efficiently with both image compression and retrieval functions. It is critical to develop image indexing techniques that search databases based on their content in a compressed domain. We propose a new scheme, query by index image , based on vector quantization, to facilitate image retrieval in a compressed domain. The proposed algorithm exploits different index images obtained by sorting codevectors to capture various kinds of image feature. Hence, intrablock correlation and interblock correlation in an image can be efficiently represented. Our proposed algorithm not only can extract features from the pixel domain but also from a transform domain, such as that of wavelet coefficients. Experimental results demonstrate that the retrieval performance of the proposed scheme is more accurate than that of other similar methods.

A model-based texture recognition system that classifies image textures seen from different distances and under different illumination directions is presented. The system works on the basis of a surface model obtained by means of four-source color photometric stereo (CPS), used to generate 2-D image textures as they would have appeared if imaged under different imaging geometries. The proposed recognition system combines cooccurrence matrices for feature extraction with a nearest neighbor classifier. The use of the cooccurrence matrices instead of filtering methods for feature extraction allows us to utilize only pixels for which valid information has been extracted by CPS. The validity of the method is demonstrated by classifying texture images captured under different imaging geometries than the reference images in the database. Moreover, the process of recognition allows one to guess the approximate direction of the illumination used to capture the test image.

A new image hiding scheme is proposed that can embed color secret images into color host images. The secret image to be embedded is first color quantized using a palette previously sorted by the principal component analysis technique to generate the indexed color image. Then the concept of pixel difference is employed to process both the host image and the indexed color image. Then the difference image of the indexed color image is embedded into the difference image of the color host image. According to our experimental results, the image quality of the stegoimage is quite good when our new method is used to embed a color secret image into the color host image.

Texture is one of many important features to capture in image characteristics. A recent texture unit-based texture spectrum approach, referred to as texture unit coding (TUC) developed by Wang and He has shown promise in texture classification. We present a new texture fea- ture extraction coding, called gradient texture unit coding (GTUC) that is based on Wang and He's texture unit to capture gradient changes in a texture unit. Since the GTUC also generates an 8-D ternary texture fea- ture vector in the same way that the TUC does, a GTUC-generated feature vector can be further represented by a number in the same range generated by the TUC. As a result, the GTUC-generated numbers also form a texture spectrum similar to that formed by the TUC-generated numbers. By normalizing a texture spectrum as a probability distribution, this work further develops an information divergence (ID)-based discrimi- nation criterion to measure the discrepancy between two texture spectra, a concept yet to be explored in texture analysis. To compare the GTUC to the TUC in texture classification, several criteria used in hyperspectral image analysis are also introduced for performance analysis. © 2004

We describe a method for generating the color texture corresponding to a range image. Based on the projection of fringe patterns on an object, a color texture of the object's surfaces can be extracted from an original fringe image. Since the range image of the object is also obtained from the original fringe image, this method eliminates alignment problems during the texture mapping, and avoids the need for extra cameras to acquire the texture. Experimental results are given to illustrate the validity of color texture extraction using our proposed method.

Morphological granulometric moments have proven useful for quantification and classification of image texture. This paper proposes using granulometries to provide a comprehensive description of surface roughness. To support this proposition it demonstrates that granulometric surface description includes much of the information carried by conventional roughness measures. It does so by using granulometric moments as inputs to a linear system to estimate six conventional surface roughness measures. The analysis is based on simulations in the framework of the Boolean-random-function model for surfaces. Features are obtained from both opening and closing granulometries. The granulometries are generated by computationally efficient one-dimensional linear and triangular structuring elements. This permits fast implementation using special-purpose algorithms. Estimators with granulometric inputs are designed on training data and applied to test data to analyze their effectiveness as estimators of conventional roughness measures. This is done for both simulated and real images.

A new stereo image compression algorithm is described in which the residual image, extracted from the stereo image by the disparity-compensated prediction method, is compressed using the wavelet transform in consideration of the inter and intra correlation between subbands. Contrasting the conventional algorithms such as efficient pyramid image code (EPIC), embedded predictive wavelet image coder (EPWIC), and JPEG, compression performance of the suggested method is significantly improved through computer simulation. Finally, it is suggested that the stereo image having a good 3-D effect can be reconstructed from the compressed image data using a new algorithm described.

We describe a novel technique for color image segmentation. This technique includes three components: a color space transformation, Markov random field expectation-maximization (MRF-EM) segmentation, and a multiresolution implementation termed ‘‘narrow band.’’ The color space transformation, from RGB to LUV, contributes to a perceptually reasonable segmentation result. The MRF-EM algorithm enables unsupervised segmentation and enforces a region smoothness constraint. The narrow-band, multiresolution implementation confines fine resolution processing to a small fraction of the image, thereby achieving acceleration over traditional multiresolution schemes. Experimental results with medical and TV images demonstrate the efficacy of the proposed technique.

The demand for high-resolution images is gradually increasing, whereas many imaging systems have been designed to enable a certain level of aliasing during image acquisition. in this sense, digital image processing approaches have recently been investigated to reconstruct a high-resolution image from aliased low-resolution images. However, since the subpixel motion information is assumed to be accurate in most conventional approaches, the satisfactory high-resolution image cannot be obtained when the subpixel motion information is inaccurate. Hence, we propose a new algorithm to reduce the distortion in the reconstructed high-resolution image due to the inaccuracy of subpixel motion information. For this purpose, we analyze the effect of inaccurate subpixel motion information on a high-resolution image reconstruction, and model it as zero-mean additive Gaussian errors added respectively to each low-resolution image. To reduce the distortion, we apply the modified multichannel image deconvolution approach to the problem. The validity of the proposed algorithm is demonstrated both theoretically and experimentally.

Using image-based metrics, contrast-to-clutter modeling is applied to the Search-2 visible image set and perception experiment data. To calculate the contrast metric, a new image is generated from the original image by replacing the target with an "expected background" using the local background surrounding the target and the natural horizontal correlation present in most surface-to-surface scenes. The contrast metric is obtained from the difference of this new image and the original image. Via a simple mathematical formula, the ratio of the contrast measure to a clutter metric is used to predict performance.

The effect of imagery wavelength on the agreement level between various image metrics and human detection performance of targets embedded in natural images is addressed. Three specially designed psychophysical experiments were conducted to record human detection performance of targets embedded in natural infrared images (3 to 5 and 8 to 12 ?m) and in natural images taken in the visual region of the spectrum. The image metrics used were designed to imitate some of the most important human perceptual cues, such as global image complexity (clutter), local target to background distinctness (contrast), and texture. The agreement level between the experimental results and the metrics products was evaluated by correlating the metrics values with target detection results obtained from each of the three psychophysical experiments. In general, it was found that the longer the imagery wavelength, the greater the dominance of global image complexity on human detection performance, and that local target to background distinctness correlates with performance for short wavelength imagery. It was also found that image texture correlates with human detection performance with very little dependence on imagery wavelength.