Search

This paper is devoted to decomposing one given image into the cartoon and texture. The cartoon parts correspond to the main large objects in the image, and the textural parts contain fine scale details. Mathematically, these two components should belong to the different function spaces corresponding to different norms. In most variational models of image decomposition, the cartoon parts are described by the total variation norm and the texture parts are characterized by its dual norm. Using these methods, the cartoon and texture can be separated from the original image. Some structures, such as edges, are well preserved in the cartoon image. However, as far as we know, none of these methods considers the geometric characteristics information of the texture position in the separated cartoon images. That is, the geometrical information in those separated cartoon images may have been destroyed, which can be seen from the experiments in this paper. To maintain these geometrical characteristics, we introduce one smoothed vector field and let it approximate to the negative gradient direction of the cartoon image, and then for the smoothed vector field, we let it belong to the Lebesgue space, thus a new variational decomposition model is proposed. The corresponding alternating direction method is discussed in detail. Experimental results are reported to show the visual qualities compared with other methods.

Image registration has always been an important research topic. This paper proposes a novel method of constructing descriptors called the histogram of oriented local binary pattern descriptor (HOLBP) for fast and robust matching. There are three new components in our algorithm. First, we redefined the gradient and angle calculation template to make it more sensitive to edge information. Second, we proposed a new construction method of the HOLBP descriptor and improved the traditional local binary pattern (LBP) computation template. Third, the principle of uniform rotation-invariant LBP was applied to add 10-dimensional gradient direction information to form a 138-dimension HOLBP descriptor vector. The experimental results showed that our method is very stable in terms of accuracy and computational time for different test images.

Contour map has contour lines that are significant in building a digital elevation model (DEM). During the digitization and preprocessing of contour maps, the contour line intersects with each other or breaks apart resulting in broken contour segments. These broken segments impose a greater risk while building DEM leading to a faulty model. The endpoints of the broken segments are matched and reconnected accurately and efficiently using the concept of minimum Euclidean distance and gradient direction. Elevation value is extracted from the map, and a digital elevation model is built in 3D space.

In this study, the authors propose a method to calculate the consistency of alpha masking to assess the robustness of the matting algorithm. This study evaluates consistent alpha masks based on the Gaussian–Hermite moment in combination with gradient amplitude and gradient direction. The gradient direction describes the appearance and shape of local objects in the image, and the gradient amplitude accurately reflects the contrast and texture changes of small details in the image. They selected Gaussian blur, pretzel noise, and combined noise to destroy the image, and then evaluated the consistency of the original alpha mask and noise alpha mask. To determine the robustness of the matting algorithm, they assessed the degree of consistency of the alpha mask using three different evaluation levels. The experimental results show that noise has a greater impact on the performance of the matting algorithm, which shows a decreasing trend as the noise level in the image deepens. In noisy images, the traditional matting algorithm exhibits better robustness compared to the recently proposed trap matting algorithm. Different matting algorithms present different adaptations to different noises.

Restoring a damaged image is a challenging topic in the field of image restoration. The famous previous method for restoring a degraded image are filters (inverse and wiener) and maximum a posteriori (MAP) formulation. However, that method has limited performance for restoring damaged images. In this paper, the multi mirroring method have been implemented for reconstructing damaged image which based on gradient direction. Firstly, the method will detect damaged image areas and then the multi mirroring method is implemented for filling a damaged image area. The simulation result shows that the proposed method has good result and capable to restore the damaged image.

In this paper, we propose a generation method of visually protected images and its application to privacy-preserving machine learning. Images generated by the proposed method hold the gradient direction information of the original images, but have no the information. Histogram-of-Oriented-Gradients (HOG) features are extracted from the protected images, and the features are applied to machine learning algorithms. In addition, the proposed generation method is an irreversible one, so there is no need to manage secret keys, unlike encryption methods. In an experiment, a face classification task is carried out under the use of a support vector machine algorithm with the HOG features to demonstrate the effectiveness of the proposed method.

The article offers a new method to detect brightness edges. It requires no image pre-smoothing and forms a contour image with minimum number of short lines. It is an analogue and, at the same time, an alternative to the widely known Canny border detection method. It differs from it by the method of calculating partial derivatives and by the choice of thresholds in the automatic mode. The processing time of one frame in the proposed method is 3 times less than in the Canny method.

International audience; Shock filter represents an important family in the field of nonlinear Partial Differential Equations (PDEs) models for image restoration and enhancement. Commonly, the smoothed second order derivative of the image assists this type of method in the deblurring mechanism. This paper presents the advantages to insert information issued of oriented half Gaussian kernels in a shock filter process. Edge directions assist to preserve contours whereas the gradient direction allow to enhance and deblur images. For this purpose, the two edge directions are extracted by the oriented half kernels, preserving and enhancing well corner points and object contours as well as small objects. The proposed approach is compared to 7 other PDE techniques, presenting its robust-ness and reliability, without creating a grainy effect around edges. Since 1960, digital images may simply be deblurred by combining the difference between an original image I 0 and ∆I: a blurred version of this same image. Usually, ∆I corresponds to a blur process equivalent to the heat equation or a convolution of I 0 with an isotropic Gaussian. This original theory proposed by Gabor is proportional to using the Laplacian operator [7]. Thus, a simplest manner to remove blur in an image remains the equation: ∂I ∂t = I 0 − α · ∆I, (1) where t represents the time or the observation scale and α < 1 is a little scalar to control the deblurring. This process is equivalent to the inverse heat equation. However, this technique is not stable because the procedure blows up after several iterations and generates an unusable image [7]. To improve eq. 1, rather than applying a global operator on all the image, the main idea is to iterate local operator at level of each pixel. Nonlinear Partial Differential Equations (PDEs) may achieve this task [7,2], practicing anisotropic diffusions of pixel information in the image. Indeed, PDEs belong to one of the most important part of mathematical analysis and are closely related to the physical world. In this context, images are considered as evolving functions of time and a regularized image can be seen as a version of the original image at a special scale.

Strawberry picking by machines confronts a complex environment where a targe may be sheltered by leaves or overlap with each other. Also, it is a challenge for machines to recognize mature strawberries among those in different maturity. This work presents a fast recognition method for elevated mature strawberries by the approach of deep learning. It uses an Ostu algorithm to separate targets from background and then the resulted effective image areas designated by the minimum external rectangular marking method are used to train CaffeNet for automatic target recognition. For comparison, we also design a SVM classifer that uses HOG gradient direction feature and H component of the color feature of the mature strawberries. The experimental results show that the average recognition rate of mature strawberries by CaffeNet can reach 95%, higher than that by SVM by 11%.

Proceedings of the 8th International Conference, ICISP 2018, Cherbourg, France, July 2-4, 2018; International audience; Easy to use, oriented half kernels are reliable in image analysis. These thin filters, rotated in all the desired directions are useful to detect edges, or extract precisely their orientations, even concerning highly noisy images. Usually, the filtering process corresponds to convolutions with Gaussians and their derivatives. Other filters exist and can be implemented in order to build half kernels. However, functions used for the smoothing and derivative parts have not been studied in depth. The goal of this paper is to evaluate different types of half filters as a function of the noise level. The studied kernels have the same spatial support, enabling easier comparisons. To address the robustness of the studied filters against noise, the image quality is gradually worsened. Then, their performances are compared through objective evaluations of both segmentation and gradient direction.

In this paper, we proposed an infrared (IR) target detection method based on the receptive field (RF) and lateral inhibition (LI). In this method, the direction parameters of Gabor filter is adaptively determined according to the gradient direction. And a background prediction method based on LI is used for regulating the gray value in image so as to achieve background suppression and target enhancement. Experimental results indicate that the proposed method can extract both small and area target from complex background, and the target detection ability is satisfactory.

Robust facial representation approach is critical for face recognition. LDP is a more stable robust descriptor using gradient direction instead of intensity value. But it is less precise to treat the directional response values in the same way and it does not obtain enough information only considering fixed absolute values of the edge responses. we propose an adaptive local directional pattern (ALDP) feature descriptor for face recognition in this paper. Positive and negative edge directions are extracted to explore more valuable discriminant information in our ALDP. Based on Weber's law, an automatic threshold setting strategy is proposed to make the ALDP codes flexible and precise. The experiment results indicate our ALDP has higher recognition accuracy in comparison with traditional methods.

In order to improve efficient of compressed sensing image reconstruction, an improved gradient projection algorithm of compressed sensing theory is proposed. In improved Gradient Projection algorithm, the pursuit direction is updated by search at negative gradient direction, thus the gradient direction is a single direction, because the traditional gradient projection algorithm searching at alternating searching method,the efficient of gradient projection algorithm is higher than the traditional gradient projection algorithm, Experiment results show that, compared with the GPSR algorithm, the IGPSR algorithm improves the signals reconstruction accuracy, improves PSNR of reconstruction signals, and exhibits higher robustness under different noise intensities. (2014) Trans Tech Publications, Switzerland.

This paper presents an effective 3D local feature descriptor, which is called Gradient Direction Accumulation-based Heat Kernel Signature (GDA-HKS) descriptor, and its application in nonrigid 3D model retrieval. The GDA-HKS descriptor is based on the heat kernel signature, and it is scale invariant and robust to the nonrigid deformation of the 3D model. Compared with the SI-HKS descriptor, the GDA-HKS descriptor is constructed directly in the time domain, and it can effectively avoid the loss of high frequency information. The absolute gradient difference is used to encode the GDA-HKS descriptor, which can describe the changing trend of the one-dimensional signal more effectively. Extensive experimental results have validated the effectiveness of the designed GDA-HKS descriptor.

In edge detection algorithms, there is a common redundancy problem, especially when the gradient direction is close to -135°, -45°, 45°, and 135°. Double edge effect appears on the edges around these directions. This is caused by the discrete calculation of non-maximum suppression. Many algorithms use edge points as feature for further task such as line extraction, curve detection, matching and recognition. Redundancy is a very important factor of algorithm speed and accuracy. We find that most edge detection algorithms have redundancy of 50% in the worst case and 0% in the best case depending on the edge direction distribution. The common redundancy rate on natural images is approximately between 15% and 20%. Based on Canny’s framework, we propose a restriction in the hysteresis step. Our experiment shows that proposed restricted hysteresis reduce the redundancy successfully. INTRO – INteractive RObotics research network

SUMMARY This paper proposes a method of accurately detecting the boundary of narrow stripes, such as lane markings, by employing gradient cues of edge points. Using gradient direction cues, the edge points at the two sides of the boundary of stripes are classified into two groups before the Hough transform is applied to extract the boundary lines. The experiments show that the proposed method improves significantly the performance in terms of the accuracy of boundary detection of narrow stripes over the con

Extracting axis of 3D blood vessel images is very important and useful to quantify blood vessel in medical diagnosis. According to blood vessel features, we constructed the energy constraint equation of blood vessel. The initial skeleton curve of blood vessel images obtained by thinning algorithm dynastically converges to the position of the axis under energy constraint equation and along gradient direction of the distance field of blood vessel images. When the equation energy reaches a minimum value, the initial skeleton curve also fixes in the axis position at this time. Experimental results show that the position of the blood vessels axis extracted by this method is accurate, and the axis preserves topology and connectivity.

According to Mallat multi-resolution analysis, A new edge detection algorithm based on multiscale product is presented, which uses Gaussian function and its first-derivative as lowpass and highpass filter to enhance edge and suppress noise, then detect edge embedded noise by gradient direction and updating search method. The experiments show that this approach has advantages of detecting edge in different gray contrast, high signal-noise ratio and pixel-level location accuracy.

In this paper a face location technology based on gradient distributions is presented. It begins with vertical location by use of vertical integral projection on the two-valued image of the original image, and then proceeds with horizontal location according to the distributions of gradient direction. Experiments have proved this technology fast and efficient.

Face recognition under changing lighting conditions is a challenging problem in computer vision. In this paper, we analyze the relative strengths of different lighting insensitive representations, and propose efficient classifier combination schemes that result in better recognition rates. We consider two experimental settings, wherein we study the performance of different algorithms with (and without) prior information on the different illumination conditions present in the scene. In both settings, we focus on the problem of having just one exemplar per person in the gallery. Based on these observations, we design algorithms for integrating the individual classifiers to capture the significant aspects of each representation. We then illustrate the performance improvement obtained through our classifier combination algorithms on the illumination subset of the PIE dataset, and on the extended Yale-B dataset. Throughout, we consider galleries with both homogenous and heterogeneous lighting conditions.

Quality control is a crucial issue in a float glass factory, and defects existing in float glass can dramatically depress glass grade. Manual inspection in float glass quality control cannot catch up with the development of float glass industry, and automatic glass defect inspection has been a trend. An online defects inspection method for float glass based on machine vision is presented in this paper, and a distributed online defect inspection system for float glass fabrication is realized. This method inspects defects through detecting the change of image gray levels caused by the difference in optic character between glass and defects. A series of image processing algorithms are set up around the analysis of glass image and the requirements of online inspection system such as reliability, real-time, and veracity. Image filtration based on gradient direction is used to filter noise and reserve the source information of defects. Downward threshold based on adaptive surface removes the background composed with stripes and strengthens defect features. Distortion part and core part of defects are obtained through fixed threshold and OTSU algorithms with gray range restricted, respectively. The fake defects (insects, dust, etc.) are eliminated based on the texture of real defects. The application of an inspection system based on this method in Wuhan glass factory proves this inspection method is effective, accurate, and reliable.

Depth image based rendering (DIBR) is one of the key techniques in converting the 2D videos into the stereoscopic 3D ones. Warping process is the major step in DIBR, which generates the virtual left and right views with a number of black disocclusion areas. The disocclusion filling process is the main task in DIBR. In this paper, an improved disocclusion filling approach is proposed, which is based on the texture and gradient analysis. Inpainting algorithm is applied first based on the background texture and depth information to repair the disocclusion areas, which can be effectively reduce the large sizes of the disocclusion areas. Then a new gradient direction template is proposed and applied to fill the left disocclusion areas. Experiments are conducted to compare with other disocclusion filling approaches. Our results are promising.

Robots are now used routinely to perform crystallization experiments and many laboratories now have imaging systems to record the results. These images must be evaluated rapidly and the results fed back into optimization procedures. Software to analyse the images is being developed; described here are methods to restrict the area of the image to be analysed in order to speed up processing. Properties of the gradient of greyscale images are used to identify first the well and then the crystallization drop for various crystallization trays and different imaging systems. Methods are discussed to identify artefacts in the images that are not related to the experimental outcome, but can cause problems for the machine-learning algorithms used in classification and waste time during analysis. Gradient angles are exploited to eliminate faults in the crystallization trays, bubbles and splatter droplets prior to analysis.

Face detection from cluttered images is challenging due to the wide variability of face appearances and the complexity of image backgrounds. This paper proposes a classification-based method for locating frontal faces in cluttered images. To improve the detection performance, we extract gradient direction features from local window images as the input of the underlying two-class classifier. The gradient direction representation provides better discrimination ability than the image intensity, and we show that the combination of gradient directionality and intensity outperforms the gradient feature alone. The underlying classifier is a polynomial neural network (PNN) on a reduced feature subspace learned by principal component analysis (PCA). The incorporation of the residual of subspace projection into the PNN was shown to improve the classification performance. The classifier is trained on samples of face and non-face images to discriminate between the two classes. The superior detection performance of the proposed method is justified in experiments on a large number of images.

In many applications, such as medical diagnosis, correctness of volume rendered images is very important. The most commonly used method for gradient calculation in these volume renderings is the Central Difference Method (CDM), due to its ease of implementation and fast computation. In this paper, artifacts from using CDM for gradient calculation in volume rendering are studied. Gradients are, in general, calculated by CDM with second-order accuracy, \(\mathscr {O}(\varDelta x^2)\). We first introduce a simple technique to find the equations for any desired order of CDM. We then compare the \(\mathscr {O}(\varDelta x^2)\), \(\mathscr {O}(\varDelta x^4)\), and \(\mathscr {O}(\varDelta x^6)\) accuracy versions, using the \(\mathscr {O}(\varDelta x^6)\) version as “ground truth”. Our results show that, unsurprisingly, \(\mathscr {O}(\varDelta x^2)\) has a greater number of errors than \(\mathscr {O}(\varDelta x^4)\), with some of those errors leading to changes in the appearance of images. In addition, we found that, in our implementation, \(\mathscr {O}(\varDelta x^2)\) and \(\mathscr {O}(\varDelta x^4)\) had virtually identical computation time. Finally, we discuss conditions where the higher-order versions may in fact produce less accurate images than the standard \(\mathscr {O}(\varDelta x^2)\). From these results, we provide guidance to software developers on choosing the appropriate CDM, based upon their use case.

This paper presents a new method of license plate location under complex background. The texture direction map was obtained by gradient direction field and the calculation of the original image. License plate candidate area is determined using the method of interval judgment edge information through texture direction and binary image. Finally, the plates are accurately positioned using the improved regional notation. This experiments results demonstrate the great robustness and efficiency of our method.

We proposed a simple and dynamic wavelet-based algorithm for enhancement of the satellite image sharpness or blurriness of a satellite image. Four set of methods are followed here (Denoising, Decomposing, Sharpness Estimation and Filtering). First Denoising is done on the input images and then it operates by initially decomposing the input image through a multilevel separable DWT. After this, the log-energies of the DWT sub bands are computed. A Scalar Index corresponding to the input image's sharpness is computed through the weighted average of the computed log-energies. Along with the Scalar Sharpness Index, a Block based algorithm is presented to determining the local perceived sharpness. This method is the simplest, fastest and accurate comparing to the currently best-performing techniques for the sharpness estimation. In this Paper A robust quantization-based image watermarking scheme, called the gradient direction watermarking (GDWM) is introduced for satellite enhanced image watermarking. In GDWM, the watermark bits are embedded by quantizing the angles of significant gradient vectors at multiple wavelet scales. To quantize the gradient direction, the DWT coefficients are modified based on the derived relationship between changes in the coefficients and the change in the gradient direction. This Watermarking technique is more robust for various sizes of watermark image. When compared to the existing watermarking techniques the proposed enhanced watermarking technique shows best results.

In this paper, a novel template matching algorithm named radial ring code histograms (RRCH) for multi-objects positioning is proposed. It is invariant to translation, rotation and illumination changes. To improve the identification ability of multi objects with different rotation angles, radial gradient codes using relative angle between gradient direction and position vector is proposed. Adjustable weights in different regions make it possible to adapt various type objects. Experiments using a LED sorting equipment demonstrate that our algorithm results in correct positioning for multi objects in complicated environments with noise and illumination invariance.

Feature extraction is an important step in facial expression recognition. A novel method is proposed based on feature fusion which combines gradient direction and LBP features. Firstly, eyes are located through the integration projection method. And the operation of image rotating, cropping and normalizing is conducted based on eyes’ position. Secondly, the image is partitioned into nine non-overlapping regions with different weight, then the gradient direction and LBP features are extracted and fused. The fused features generated from each of the regions are concatenated to form the feature vector which represents the facial expression. Finally, K-nearest neighbor algorithm is performed for classification. Experiments on JAFFE and Cohn-Kanade facial expression databases show that the proposed method achieves better performance for facial expression recognition.

AUV localization is not accurate based on sequence images if moving target is as landmark, so the moving target detection algorithm is studied based on global motion estimation, which detects and eliminates moving target according to the motion inconsistency of the moving target. Generally grid block matching is used in the global motion estimation, it can’t effectively dispose the dynamic background, and the gradient direction invariant moments descriptors method of free circular neighborhood based on feature points is proposed, which is effective for the background rotating and light changing in two adjacent images. For the matching points, the parameters of global motion are estimated robustly combined with normalized linear estimation method and least median squares. Experiments show that the designed algorithm can effectively estimate parameters of global motion, and eliminate the motion target as mismatch.

An image mosaic method that can be used when creating advertisements or posters is proposed in this study. Mosaic is a method that expresses an entire image using an arbitrary number of cells. Photomosaic generates new images using a combination of photos. In this paper, we propose a new mosaic algorithm that generates an abstract artistic mosaic image by filling a region that is divided by a boundary using a unit image, which is an image that only has a shape and no allocated color. A unit image can be changed diversely through rotation or shifting, and the corresponding region is filled by using the gradient direction and edge information of the input image. For this, we extract and use information from input image such as color, edge and gradient. In result we can generate various abstractive images which can be used in advertisement and multimedia contents market.

In this paper, we propose a new angle quantization index modulation (AQIM) method, called the difference AQIM (DAQIM) method. The proposed method aims to improve the watermarking performance against gain attacks. Unlike the original AQIM method [1], the DAQIM method quantizes the difference of the two angles instead of the angles themselves. The experimental results demonstrate that our proposed method outperforms both the AQIM and the gradient direction watermarking (GDWM) [2] methods in terms of document-to-watermark ratio (DWR).

Depth images have advantages of simple processing, fog penetration, and little affection by light, thus a body shape detection algorithm based on depth image was proposed to judge personnel evacuation. This study started by making body shape dataset using a depth sensor, then extracting the HOG-depth feature. The best parameters were found, including the range of gradient direction and the number of bins. Next step was to train and classify the body shape dataset using different classifiers, and gentle Adaboost algorithm based on CART weak classifiers got the best result. Then we discussed the effect of traversal method of sliding window, and found a better pixel number of every moving step. At last, the intellectualized control method under actual personnel evacuating situation was completed from the view of software implementation.

Text information in natural scene images is important for various kinds of applications. In this paper a novel method based on stroke width to detect text in unconstrained natural scene images is proposed. Firstly, we use the stroke width transform to generate a rough estimation of stroke width map, then use K-Means clustering and the elbow method to find some specific stroke width values that are both dominant and consistent. Secondly, in order to generate better edge detection and gradient direction results we use these specific stroke width values as the size parameters in the superpixel algorithm to generate smooth and uniform region boundaries. Finally, we try to refine the stroke width map and recover valid edge pixels by applying stroke width regularized constraints on the improved edge detection and gradient direction results computed from these region boundaries. Our method was evaluated on three benchmark datasets: ICDAR 2005, 2011 and 2013, and the experimental results show that it achieves state-of-the-art performance.

Image edge gradient direction not only contains important information of the shape, but also has a simple, lower complexity characteristic. Considering that the edge gradient direction histograms and edge direction autocorrelogram do not have the rotation invariance, we put forward the image retrieval algorithm which is based on edge gradient orientation statistical code (hereinafter referred to as EGOSC) by sharing the application of the statistics method in the edge direction of the chain code in eight neighborhoods to the statistics of the edge gradient direction. Firstly, we construct then-direction vector and make maximal summation restriction on EGOSC to make sure this algorithm is invariable for rotation effectively. Then, we use Euclidean distance of edge gradient direction entropy to measure shape similarity, so that this method is not sensitive to scaling, color, and illumination change. The experimental results and the algorithm analysis demonstrate that the algorithm can be used for content-based image retrieval and has good retrieval results.

In this paper, we consider the problem of removing jaggy artifacts from images. We consider the kernel regression framework and propose a reduced-rank quadratic adaptive method that adapts to the local gradient direction. The proposed technique is effective in shrinking isophote fluctuations, and the result is smooth edges. We observe that it is critical to differentiate jaggy artifacts from texture, junctions and corners, so that meaningful image structure is preserved. Here, we demonstrate that the spectrum of the local covariance matrix of gradients, also known as the structure tensor, is well suited for differentiation of jaggy artifacts from image structure, and we incorporate this into the kernel regression framework. Results show the efficacy of the approach. Namely, that the method is effective in reducing jaggy artifacts without blurring meaningful image structure.

In this paper, we present a novel family of multiscale local feature descriptors, a theoretically and intuitively well justified variant of SURF which is straightforward to implement but which nevertheless is capable of demonstrably better performance with comparable computational cost. Our family of descriptors, called Gauge-SURF (G-SURF), is based on second-order multiscale gauge derivatives. While the standard derivatives used to build a SURF descriptor are all relative to a single chosen orientation, gauge derivatives are evaluated relative to the gradient direction at every pixel. Like standard SURF descriptors, G-SURF descriptors are fast to compute due to the use of integral images, but have extra matching robustness due to the extra invariance offered by gauge derivatives. We present extensive experimental image matching results on the Mikolajczyk and Schmid dataset which show the clear advantages of our family of descriptors against first-order local derivatives based descriptors such as: SURF, Modified-SURF (M-SURF) and SIFT, in both standard and upright forms. In addition, we also show experimental results on large-scale 3D Structure from Motion (SfM) and visual categorization applications.

Kertas kerja ini dikemukakan untuk menjelaskan tentang pembetulan pencongan dan erotan imej jawi. Satu algoritma mudah dan pantas ini menggunakan histogram orientasi cerunan sama seperti yang telah dicadangkan oleh Sun dan Si(1997) terhadap teks Latin telah dicadangkan untuk mengatasi masalah pencongan dan erotan terhadap teks Jawi. Algoritma bermula dengan melaksanakan operasi cerunan imej; diikut oleh penentuan histogram orientasi dan licinkannya menggunakan penapis median. Nilai maksimum histogram dicari untuk mendapatkan sudut pencong; diikuti oleh pembetulan awal dengan menggunakan fungsi polinominal. Sudut dihitung secara analitik; dan akhir sekali imej diputarkan mengikut sudut pencong yang diperoleh. Erotan teks Jawi juga telah menggunakan kaedah yang sama untuk mendapatkan sudut erotan dan kemudian diricih supaya dapat menghasilkan imej yang betul. Hasil daripada algoritma ini menjanjikan keputusan yang baik. Kata kunci: Histogram orientasi cerunan; topeng penjejak sisi sobel; pencong; erotan; penapis median; dan sudut pencong This paper discusses an enhancement algorithm of skew and slant orientation of the Jawi images. A simple and quick algorithm that uses a gradient orientation histogram, similar to the one proposed by Sun and Si(1997) for the Latin text is suggested to overcome this skew and slant problem in Jawi text. This algorithm start by carrying out the image gradient operation, followed by determining the orientation histogram and smoothing algorithm with a median filter. The histogram´s maximum value is generated to obtain the skew angle followed by the correcting procedure of the early value obtained by using polynomial function. The angle is calculated analytically and finally converting the image accordingly. The slanted Jawi text uses the same approach as the above to get the slant angle. This image is sheared in order to get the righ image. The outputs of this study show promising and convincing results. Key words: Gradient orientation histogram; the mask of sobel edge detector; skew; slant; median filter; and skew angle

A lane mark detection is used in various driver assist systems for vehicle electronic technology. One of the subjects for the vision system is improvement of robustness. Various methods have been tried to achieve it. In this work, we tried to improve the approach for the robustness. Our main idea is the noise reduction based on narrowing a width of search area. The proposed method uses the road model based on the generalized hough transformation. We use the gradient direction to reduce the number of votes and the method is Kernel-based Hough transform. It is shown that the method can achieve better detection result.

An algorithm based on oriented centroid for image matching is designed. The main contribution lies in providing a stable feature direction of pixels called oriented centroid which takes place of the main gradient direction of the traditional SIFT (Scale Invariant Feature Transform) algorithm. The computation cost required to calculate the oriented centroid based feature direction is much less than that required to estimated the main gradient direction, leading to the considerable improvement of the algorithm's real-time performance. Experimental results verify that the algorithm does not lose performance in terms of robustness and accuracy compared with the standard SIFT algorithm.

Edge detection with sub pixel accuracy is a vital step of vision measurement technology. In this paper, a novel sub-pixel edge detecting algorithm based on orthogonal moment is proposed. The proposed approach locates the edge points with different method for different edges. For the ideal edges, it uses the orthogonal moment based method, and for the blurred edges, the centroid based method is used. Experimental results demonstrate that the proposed method prevails in robustness and high precision compared with traditional method.

Face alignment aims to fit a deformable landmark-based mesh to a facial image so that all facial features can be located accurately. In discriminative face alignment, an alignment score function, which is treated as the appearance model, is learned such that moving along its gradient direction can improve the alignment. This paper proposes a new face model named “Optimal Gradient Pursuit Model”, where the objective is to minimize the angle between the gradient direction and the vector pointing toward the ground-truth shape parameter. We formulate an iterative approach to solve this minimization problem. With extensive experiments in generic face alignment, we show that our model improves the alignment accuracy and speed compared to the state-of-the-art discriminative face alignment approach.

In this paper a fast face location method based on gradient distributions is presented. This method can locate face with certain rotation, or light variation or with glasses or beard in the image. It begins with vertical location by use of vertical integral projection on the two-valued image of the original image, and then proceeds with horizontal location according to the distributions of gradient direction. Experiments have proved this method fast and efficient.

The existing iris location algorithms are of low executing speed and poor robustness. In order to improve the accuracy of iris location, an iris location algorithm based on gradient direction information is proposed according to both the ring structural characteristics of iris and the gray distribution features of eye image in this paper. Experiments shows that the algorithm is effective and can improve the speed and accuracy of iris location.