Image registration is a challenging task, with applications in surveillance, motion estimation, and fusion systems. Due to the diversity of sensors, local distortions and large image size, satellite images are often difficult to accurately register. In the literature, local descriptor-based processing techniques, such as scale-invariant feature transforms (SIAdaptive Block ProcessingFTs), have been applied to register satellite images, which provide robust features. However, these techniques suffer from a high-computational cost, lack of features, and low-distribution quality, which affect the registration accuracy. In this paper, we develop an algorithm to register satellite images based on adaptive block processing to increase the number of features and to improve the distribution quality. In addition, outlier removal using statistical masks are associated with classical random sample consensus (RANSAC); a subsequent comparative analysis demonstrates the accuracy of the proposed method. Typically, a classical SIFT prevents its wide application in recent remote sensing, although this is no longer the case with the proposed adaptive block processing method.

Shape from focus (SFF), which relies on image focus as a cue within sequenced images, represents a passive technique in recovering object shapes in scenes. Although numerous methods have been recently proposed, less attention has been paid to particular factors affecting them. In regard to SFF, one such critical factor impacting system application is the total number of images. A large data set requires a huge amount of computation power, whereas decreasing the number of images causes shape reconstruction to be crude and erroneous. The total number of images is inversely proportional to interframe distance or sampling step size. In this paper, interframe distance (or sampling step size) criteria for SFF systems have been formulated. In particular, light ray focusing is approximated by the use of a Gaussian beam followed by the formulation of a sampling expression using Nyquist sampling. Consequently, a fitting function for focus curves is also obtained. Experiments are performed on simulated and real objects to validate the proposed schemes.

Mostly, shape-from-focus algorithms use local averaging using a fixed rectangle window to enhance the initial focus volume. In this linear filtering, the window size affects the accuracy of the depth map. A small window is unable to suppress the noise properly, whereas a large window oversmoothes the object shape. Moreover, the use of any window size smoothes focus values uniformly. Consequently, an erroneous depth map is obtained. In this paper, we suggest the use of iterative 3-D anisotropic nonlinear diffusion filtering (ANDF) to enhance the image focus volume. In contrast to linear filtering, ANDF utilizes the local structure of the focus values to suppress the noise while preserving edges. The proposed scheme is tested using image sequences of synthetic and real objects, and results have demonstrated its effectiveness.

Although numerous methods have been recently proposed in recovering shape from image focus, all follow similar precedent of focus measure application after image acquisition and depth is recovered by maximizing the focus curves obtained via this application. This is followed by utilizing approximation techniques to refine the initial depth map. In this paper we have followed an unorthodox approach to recover object shapes using image focus as cue, not following the conventional manner. At first the focused image is obtained, pursued by computing depth along the edges and their neighbors present in scene. Empty spaces in the final depth map are then calculated by surface interpolation. Experimentation and results are then discussed to support the proposed scheme.

This paper describes a novel nodule detection method that enhances false positive reduction. Lung region is extracted from CT image sequence using adaptive thresholding and 18-connectedness voxel labelling. In the extracted lung region, nodule candidates are detected using adaptive multiple thresholding and rule based classifier. After that, we extract the 3D and 2D features from nodule candidates. The nodule candidates are then classified using genetic programming (GP) based classifier. In this work, a new fitness function is proposed to generate optimal adaptive classifier. We tested the proposed algorithm by using Lung Imaging Database Consortium (LIDC) database of National Cancer Institute (NCI). The classifier was trained and evaluated using two independent dataset and whole dataset. The proposed method reduced the false positives in nodule candidates and achieved 92% detection rate with 6.5 false positives per scan.

Generally, impulse noise filtering schemes consider all pixels within a large neighborhood. However, the estimate from all pixels within the neighborhood may not be accurate. Moreover, large window may remove edges and fine details. In contrast to this approach, we propose iterative impulse noise removal scheme that emphasizes on few noise-free pixels within a small neighborhood. This iterative process continues until all noisy pixels are replaced with the estimated values. To estimate the optimal value of noisy pixel, we developed genetic programming (GP) based estimator using noise-free pixels. The estimator is constituent of useful local pixels information. Experimental results show that the proposed scheme is capable of removing impulse noise effectively while preserving the fine details. Especially, our approach has shown effectiveness against high impulse noise density.

In the past recent years, shape recovery using image focus has been investigated a lot. Traditionally, generalized Gaussian models are used to approximate the focus curves obtained after applying the focus measure. In this paper, we have proposed a new approximation model for the focus curve, using Lorentzian-Cauchy function. The focusing of light rays, by the imaging device, are approximated by using a Gaussian beam, which has the intensity profile along the propagation axis defined by Lorentzian-Cauchy function. Experiments performed on the simulated and real objects have shown the effectiveness of the proposed scheme.

The use of intelligent and sophisticated approaches in the domain of computer vision and pattern recognition is consistently increasing. This paper introduces a novel machine learning based approach for Shape From Focus (SFF), where the in-focus pixels are selected from a sequence of images. In contrast to computing focus values directly in spatial or transform domain, the proposed method first nonlinearly transforms the input space into feature space and then computes the focus value by transforming the data into eigenspace. First, the nonlinear transformation is performed by using kernel function and then Principal Component Analysis (PCA) is applied. This idea is also supported by the fact that out-of-focus is analogous to blurring and is a nonlinear phenomenon. An initial depth map is computed by maximizing the focus measure. To further refine the 3D shape, bilateral filter is applied. The proposed method is experimented using synthetic and real image sequences. The results demonstrate the effectiveness and the robustness of the new method.

Estimating 3D shape of the object is an important research topic in the area of computer vision, with a wide range of applications. This paper introduces a new method for focused-based passive methods (like SFF) to approximate the 3D shape of the object using Bezier surface. The discrete nature of image sampling results in the loss of information between two consecutive images. Conventional approximation methods optimize or interpolate focus values. We have suggested interpolating depth values instead of focus value over a small patch on the object surface. The method approximates the surface more accurately and also reduces any noise caused by focus measure. The proposed method is tested and analyzed to demonstrate the effectiveness against traditional methods.

In conventional focus measures, focus values are locally aggregated to suppress the noise and to obtain better depth maps. However, this enlarges the difference between focus values of two consecutive frames which results in inaccurate shape. In this paper, we propose a nonparametric approach for 3D shape from image focus by applying an unsupervised formulation of kernel regression estimate. The focus volume is obtained through a focus measure and then Nadaraya and Watson Estimate (NWE) is applied to each frame. The depth is then computed by finding the frame number which maximizes the focus value. The kernel regression is again applied on depth values to obtain an accurate 3D shape. The proposed approach is experimented using synthetic and real image sequences. The results demonstrate the effectiveness of the proposed approach.

This paper presents a fast algorithm for H.264 fractional motion estimation (ME). In H.264 ME is the most time consuming component. The ME process consists of two stages: integer pixel and fractional pixel search. To reduce the complexity of fractional pixel ME we propose a quadrant based directional fractional pixel ME algorithm that is based on the unimodal property of the fractional pixel error surface. Better computation reduction has been achieved by using this strategy. Experimental results show that as compared to fast sub-pel ME proposed in H.264, the proposed method can speedup fractional ME, with a negligible degradation in video quality.

The problem of estimating the three-dimensional (3D) geometry of an object from a sequence of images obtained at different focus settings is called shape from focus (SFF). The conventional SFF methods apply focus measure operator at each pixel using neighboring pixels in the same image frame. However, for an object with complex geometry, such methods cannot compute accurate focus level of a pixel, since the neighboring pixels in an image, taken with small depth of field, do not have the same focus level. In this paper, a novel SFF algorithm based on combinatorial optimization is proposed. After the rough estimate of the shape, we refine the shape iteratively by searching the optimal focus measure of each pixel using neighboring pixels on various image frames. The proposed SFF algorithm shows improvements in both the accuracy of the shape and the computational complexity in comparison to the previous SFF methods.

New polymers are frequently appearing in the literature with a wide range of applications, e.g., they are being used in applications like fuel cell membrane, fertilizers, coating elements, etc. One of the requirements of the researchers involved in polymer science is to quickly compare the samples being fabricated. One of the measures for comparison is surface roughness. However, the high cost involved in the estimation of surface roughness limits its extensive and exhaustive usage. In this paper, we propose to estimate the surface roughness using the image focus, i.e., shape from focus (SFF). We show that the results from SFF are encouraging for comparison of polymers using surface roughness.

This paper introduces a new robust algorithm for shape from focus (SFF). Principal component analysis (PCA) is applied to transform the data into eigenspace and the first feature is employed to calculate the depth value. Contrary to computing the focus value locally by a focus measure in first step and then, in second step, approximating the depth map, the proposed method finds the location of the best focused value over a sequence of pixels. The proposed method is experimented using synthetic and real image sequences. The evaluation is gauged on the basis of unimodality and monotonicity of the focus curve. Two other global statistical metrics root mean square error (RMSE) and correlation have also been applied for synthetic image sequence. Experimental results have demonstrated the effectiveness and the robustness of the new method.

In this paper we present a method for depth estimation using image focus for 3D shape recovery. The method is based on the regression model. First the sequence of images is processed using Laplacian operator and the first estimate for the maximum value is found. Two datasets are selected for each pixel based on that maximum value. For each of the two datasets, a straight line is approximated to fit each of the dataset. Least squares method is used to find the best fit. The intersection point of the two lines is computed and weights are used so that the final depth map can be obtained. Finally, averaging filter is used to remove noise. The proposed method is compared with earlier depth estimation algorithms using various test objects.

In the area of machine vision, shadows occur frequently in a wide variety of scenes. An inherent weakness of focus-based methods is that they require the imaged scene to have significant texture. In many real-world applications, surfaces can be smoothly shaded and lacking in detectable texture. In such cases, Shape-Form-Focus (SFF) generates inaccurate and sparse depth maps. For accurate calculation of depth map, we consider the affects of illumination. In SFF the images are taken with one light source and the shadows occur in each frame. Thus, the more bright regions of the images give more accurate depth map, whereas, the less bright regions give less accurate depth map. We have also shown the results and compared them with the previous results. From simulation results we see that the results improved with the proposed algorithm.

There are many factors affecting the depth estimation for 3D shape recovery using passive optical methods. In this paper, we consider the effects of noise, source illumination and texture reflectance for shape from focus technique. We present a focus measure which shows consistent performance for varying noise levels, source illumination levels and different texture reflectance. The focus measure is based on an optical transfer function implemented in the Fourier domain and its results are compared with four other focus measures. The additive Gaussian noise is considered for noise analysis. Three illumination levels are considered for source illumination and three different textures are studied for reflectance analysis.

Hydrophobicity has been considered as the potential measurement for the prediction of G-Proteins coupled receptor subfamilies. In the present work, using Hydrophobicity measure, we make use of fast Fourier transform to better analyze the sequence information. In our experiments, we have observed that sequence pattern based information could easily be exploited in the frequency domain using proximity rather than increasing margin of separation between the classes. Based on this information, a simple nearest neighbor (NN) method is then used to classify the 17 subfamilies. The proposed proximity based approach has outperformed the one against all implementation of support vector machine (SVM) [Y. Z. Guo, et al, Acta Biochimica et Biophysica Sinica, 37(2005) 759]. Our simple proximity based approach has superior performance in terms of all three measures on both Jackknife and independent data set. For B, C, D and F subfamilies, the Mathew's correlation coefficient and overall accuracy using jackknife test are 0.96 and 96.03%, while, using independent data set are 0.91 and 91.6% respectively. The results validate the idea of exploiting sequence pattern based information in the frequency domain using proximity in terms of Euclidian distance. Another side advantage is that instead of training and saving 17 SVM models, we need a single NN classifier.

Accurate estimation of depth map leads to precise three-dimensional shape recovery. In this paper, we present a new focus measure for calculation of depth map. This new focus measure is based on an optical transfer function implemented in the frequency domain and it has shown robustness in the presence of noise as compared to the earlier focus measures. The results of the proposed focus measure have shown considerable improvement in the presence of noise with respect to other focus measures.

This paper suggests a simple scheme in which the search pattern is divided into a number of sectors based on the spatio-temporal correlation information. In the first step five neighboring blocks are searched for finding the predicted motion vector. The predictive motion vector thus obtained is chosen as the initial search center. This predictive search center is found to be closer to the global minimum and thus decreases the effects of the monotonic error surface assumption and its impact on the motion field estimates. Secondly the prediction information is used to obtain the direction of predicted motion vector. Based on the direction of predicted motion vector the search area is divided into four sectors. Final search pattern is adaptive and depends on the sector selected and significantly reduces the computational complexity. Experiments show the speed improvement of the proposed algorithm as compared to other fast search algorithms, in addition the image quality measured in terms of PSNR also shows good results.

Three dimensional (3D) shape recovery using shape from focus (SFF) is presented as an optimization problem, i.e., maximization of the focus measure in the 3D image volume. The whole image volume (sequence) is divided into a number of sub image volumes. The search of 3D shape is made in the sub image volumes using dynamic programming optimization technique. The final depth map is obtained by collecting the depth map of the sub volumes. The new algorithm has considerably decreased the computational complexity by searching the 3D shape in sub image volumes and has shown better results. New 3D focus measure operators are also introduced for higher accuracy at the cost of some computational costs.

This paper presents a new linked significant tree (LST) wavelet coding method for efficient compression of images. The proposed method links all the significant coefficients together within a wavelet tree to facilitate encoding algorithm. Insignificant parents with at least one significant child found in the wavelet data are changed to significant to link them together. Their location is saved for identification, and they are then treated just like other significant coefficients during encoding. LST encoding process is performed next on the processed output for compression. It exploits the fact that since all the significant coefficients within a wavelet tree are linked together, no descendent of an isolated coefficient can be significant any more. This eliminates the need to check descendents of those coefficients that are found as insignificant. Locations of converted coordinates are sent to the decoder after each encoding pass, in order to change these values back to original during reconstruction. Better compression is achieved when the proposed method is applied to smooth images.

The conventional shape from focus (SFF) method for 3D shape recovery from image focus is fast but inaccurate. The SFF method based on the focused image surface (FIS) has shown better results by exhaustive search of the FIS shape using planar surface approximation at the cost of considerably higher computations. We investigate a fast and accurate SFF method. The conventional SFF method is used as a rough estimate for pixels at regular steps in the x and y directions, and this rough estimate is used to search the FIS shape for all pixels between the steps using a dynamic programming optimization technique. The proposed algorithm is very fast and shows comparable results with those of accurate SFF methods.

We propose a new method for depth from defocus (DFD) using wavelet transform. Most of the existing DFD methods use inverse filtering to determine the measure of defocus. These methods suffer from inaccuracies in finding the frequency domain representation due to windowing and border effects. The proposed method uses wavelets that allow performing both the local analysis and windowing with variable-sized regions for images with varying textural properties. We show that normalized image ratio of wavelet power by Parseval's theorem is closely related to blur parameter and depth. Experimental results have been presented demonstrating that the proposed DFD method is faster and generates more accurate depth maps than the previous methods.

The most popular shape from focus (SFF) methods in the literature are based on the concept of focused image surface (FIS)-the surface formed by the best focus points. According to paraxial-geometric optics, there is one-to-one correspondence between the shape of an object and the shape of its FIS. Therefore, the problem of three-dimensional (3-D) shape recovery from image focus can be described as the problem of determining the shape of the FIS. The conventional SFF method is inaccurate because of piecewise constant approximation of the FIS. The SFF method based on the FIS has shown better results by exhaustive search of the FIS shape using planar surface approximation at the cost of considerably higher computations. In this paper, search of the FIS shape is presented as an optimization problem, i.e., maximization of the focus measure in the 3-D image volume. The proposed method searches the optimal focus measure in the whole image volume, instead of the small volume as adopted in previous methods. The dynamic programming, instead of the approximation techniques, is used to search the optimal FIS shape. A direct application of dynamic programming on a 3-D data is impractical, because of higher computational complexity. Therefore a fast heuristic model based on dynamic programming is proposed for the search of FIS shape. The shape recovery results of the new method are better than previous methods. The proposed algorithm is significantly faster than the FIS algorithm, but a little slower than the conventional algorithm.

In this paper we propose a new watermarking method based on scene segmentation for copy protection in hard disk drive embedded digital broadcast receivers. First, video sequence is segmented as scenes using the macroblock types of B-picture in the MPEG compressed domain. Second, for each scene, different embedding parameters are determined from the image complexity and motion vector amplitude. For copy protection, copy control watermark information is embedded in the content whether the content is copyrighted or not. By using this method, invisibility and robustness of watermark, which are the basic requirements of watermarks, are satisfied at the same time.

This paper presents a novel preprocessing technique of wavelet transform data to link all the significant coefficients together, in order to facilitate the image coding algorithms for increased compression. The proposed algorithm finds the isolated significant coefficients in the wavelet transformed data for current threshold value. All the insignificant coefficients in the wavelet tree that lies above that isolated significant coefficient (i.e., its insignificant parents) are changed to significant coefficients and their location is saved, they are then treated just like other significant coefficients. Coding algorithms have been modified to accommodate the processed output. In the end, algorithm indicates converted coordinates that are used to change the value back to original during reconstruction. Noticeably high compression ratio is achieved for most of the images, when the proposed method is used with the modified image codecs.

Image indexing based on a modified color cooccurrence matrix (MCCM) is proposed. First, the CCM is simplified to represent the number of color (hue) pairs between adjacent pixels in the image. Then, the CCM is split into diagonal and non-diagonal elements that constitute two elements of the MCCM. Indexing the image by MCCM could exploit shape information in the abstract level. The proposed MCCM is an accumulative feature. Experimental results show the superiority of MCCM based indexing in comparison to indexing based on other accumulative features, color histogram and auto-correlogram, with very competitive computational cost.

Image indexing based on modified color co-occurrence matrix (MCCM) is proposed in this paper. First, CCM is simplified to represent the number of color (hue) pairs between adjacent pixels in the image. And then, CCM is split into diagonal and non-diagonal elements that constitute two elements of MCCM. Indexing the image by MCCM could exploit shape information in abstract level. Proposed MCCM is accumulative feature. Experimental results show the superiority of the proposed MCCM based indexing in comparison to the indexing based on other accumulative features, color histogram and auto-correlogram with very competitive computational cost.

A new motion estimation algorithm on the basis of a dynamic programming optimization operator (DPOO) is proposed. Motion estimation computation is formulated as a matching optimization problem of multiple dynamic images. A new operator that is a modification of dynamic programming recursion has been designed. This operator allows multiple implementations, and extends 1D optimization of the dynamic programming method to N-D optimization. Discrete Fourier transform based data level reduction for the motion estimation algorithm has been realized.

The classical color histogram for image indexing does not take into account the shape information of an image. A color histogram method with edge information is studied. Color distributions are found for the pixels of three types of edges (two directional edges and one non-directional edge) and three distance measures are computed on the basis of the color distribution of each edge type. The proposed similarity measure obtained by combining these three distance measures could reduce the false match rate in comparison with using only single (non-directional) edge types. Simulation results show an improvement in indexing quality as compared to that of the traditional color histogram and edge histogram.

The conventional shape-from-focus (SFF) methods have inaccuracies because of piecewise constant approximation of the focused image surface (FIS). We propose a scheme for SFF based on representation of three-dimensional (3-D) FIS in terms of neural network weights. The neural networks are trained to learn the shape of the FIS that maximizes the focus measure.

A new method for computing precise depth map estimates of 3D shape of a moving object is proposed. 3D shape recovery in motion stereo is formulated as a matching optimization problem of multiple stereo images. The proposed method is a heuristic modification of dynamic programming applied to two-dimensional optimization problem. 3D shape recovery using real motion stereo images demonstrates a good performance of the algorithm in terms of reconstruction accuracy.

Accurate motion estimation is the key to the next generation of video compression techniques. As part of a larger motion compensated system, accurate motion vectors will be required if the compression is going to achieve the high quality that applications are striving for. Even low bit rate coding schemes such as those used for video conferencing will be demanding better quality. Block matching algorithms that predict the motion field using blocks of pixels use search heuristics to find the best match. The full search technique will always be better at minimizing the prediction error compared to reduced search techniques. However, blocks on the boundary of moving objects create problems for accurately predicting the motion vectors, which are very sensitive for human eyes. We proposed a new method for accurately determining the motion vectors using the edge matching criteria. This method gives good motion estimation, and the motion compensated frames are very close to the original frames. This new method reduces the computational requirements significantly and gives good results as compared to the intensity based full search algorithm.

Previous studies show that the motion vector distribution within the search window shows a center biased behavior. Based on this fact, an advanced center biased search algorithm for fast block motion estimation has been proposed. The algorithm drastically reduces the computational complexity by strict application of the unimodal error surface assumption. Improved error performance has been achieved by an efficient center biased search strategy that improves the chances of getting the correct motion vector. The half stop technique has been adopted to speed up the block matching process. Experimental results show that proposed algorithm has improved performance as compared to the three step search algorithm. A good predicted image quality is also achieved.

A new approach to 3-D profilometry for the white light interferometer (WLI) is presented. We detail a simple way to construct a profiler that uses the fringe contrast degradation of the white light Mirau interferometer. The amount of fringe contrast degradation in the Mirau interferometer can be founded by the depth from focus (DFF) method. In the method, the variance of the mismatch function and the modified local variance function are used as the focus measures. The proposed method is named as the extended depth from focus (EDFF) that determines the zero optical path difference (OPD) from the amount of fringe contrast degradation of the white light interferometer. The method has a theoretically unlimited range and can profile with subpixel accuracy both optically rough and smooth surfaces without changing the algorithm.

A full-search based block-matching algorithm for motion estimation has a major problem of significant computational load. To solve this problem, extensive research in fast-motion estimation algorithms have been carried out. However, most of them have some degradation in the predicted image from the reduced computation. To decrease the amount of significant computation of the full-search algorithm, we propose a fast block-matching algorithm based on an adaptive matching scan and representative pixels without any degradation of the predicted image. By using Taylor series expansion, we obtain the representative pixels and show that the block-matching errors from the reference block and candidate blocks are proportional to the block complexity. With the derived result, we propose a fast full-search algorithm with adaptive scan direction in block matching. Experimentally, our proposed algorithm is very efficient in terms of computational speedup, and is the fastest among all the conventional full-search algorithms. Therefore, our algorithm is useful in VLSI implementation of video encoders for real-time encoding.

A new method for the three-dimensional shape recovery from image focus is proposed. The method is based on an approximation of the focussed image surface (FIS) by a piecewise curved surface which tracks the realistic FIS in image space. The piecewise curved surface is estimated by interpolation using the Lagrangian polynomial. The new method has been implemented on a prototype camera system. The experiments and their results are provided and discussed. The experimental results show that the new method gives more accurate results than the previous methods.