Zhitao Xiao

Recognition for woven fabric pattern based on gradient histogram

To identify woven fabric pattern, a robust recognition method is presented in this paper. Firstly, using gray image of woven fabric through gray projection, we correct the warp deviation and segment crossed-area of warp and weft. Then, based on the gradient histogram information, we get the texture features of crossed-area, which can determine the state of crossed-area. Through which we can obtain the recognition preliminarily, which can overcome the effects of scale and color change of woven fabric and achieve high accuracy. According to the woven fabric periodicity, the error recognition is corrected. This method is validated with real woven fabric images, which can get woven fabric pattern diagram successfully. According to the extensive tests, we concluded that this method can achieve the recognition of woven fabrics with complex patterns and colors.

Anisotropic coupled diffusion filter and binarization for the electronic speckle pattern interferometry fringes.

In this paper novel approaches based on anisotropic coupled diffusion equations are presented to do filter and binarization for ESPI fringes. An advantageous characteristic associated with the proposed technique is that diffusion takes place mainly along the direction of the edge. Therefore, the proposed anisotropic coupled diffusion filter method can avoid blur of the fringe edge and protect the useful information of the fringe patterns. The anisotropic coupled diffusion binarization, which can repair the image boundary anisotropically, is able to avoid the redundant burr. More important, it can be directly applied to the noisy ESPI fringe patterns without much preprocessing, which is a significant advance in fringe analysis for ESPI. The effective of the proposed methods are tested by means of the computer-simulated and experimentally obtained fringe patterns, respectively.

Diabetic Retinopathy Fundus Image Processing Based on Phase Information

Precise fundus image features detection is an important factor for screening diabetic retinopathy. Some noises in fundus image features extraction need to be solved. This paper studies using phase information to attempt get better effect. We use and compare four phase-based approaches and get some instructive results.

Hard Exudates Detection Method Based on Background-Estimation

Hard exudates (HEs) are one kind of the most important symptoms of Diabetic Retinopathy (DR). A new method based on background-estimation for hard exudates detection is presented. Firstly, through background-estimation, foreground map containing all bright objects is acquired. We use the edge information based on Kirsch operator to obtain HE candidates, and then we remove the optic disc. Finally, the shape features, histogram statistic features and phase features of the HE candidates are extracted. We use the SVM classifier to acquire the accurate extraction of HEs. The proposed method has been demonstrated on the public databases of DIARETDB1 and HEI-MED. The experiment results show that the method’s sensitivity is 97.3 % and the specificity is 90 % at the image level, and the mean sensitivity is 84.6 % and the mean predictive value is 94.4 % at the lesion level.

Automatic non-proliferative diabetic retinopathy screening system based on color fundus image

BackgroundNon-proliferative diabetic retinopathy is the early stage of diabetic retinopathy. Automatic detection of non-proliferative diabetic retinopathy is significant for clinical diagnosis, early screening and course progression of patients.MethodsThis paper introduces the design and implementation of an automatic system for screening non-proliferative diabetic retinopathy based on color fundus images. Firstly, the fundus structures, including blood vessels, optic disc and macula, are extracted and located, respectively. In particular, a new optic disc localization method using parabolic fitting is proposed based on the physiological structure characteristics of optic disc and blood vessels. Then, early lesions, such as microaneurysms, hemorrhages and hard exudates, are detected based on their respective characteristics. An equivalent optical model simulating human eyes is designed based on the anatomical structure of retina. Main structures and early lesions are reconstructed in the 3D space for better visualization. Finally, the severity of each image is evaluated based on the international criteria of diabetic retinopathy.ResultsThe system has been tested on public databases and images from hospitals. Experimental results demonstrate that the proposed system achieves high accuracy for main structures and early lesions detection. The results of severity classification for non-proliferative diabetic retinopathy are also accurate and suitable.ConclusionsOur system can assist ophthalmologists for clinical diagnosis, automatic screening and course progression of patients.

Automatic recognition for striped woven fabric pattern

Three basic weaves and fancy woven fabrics can be recognized usually. But the recognition of the striped woven fabric pattern is a challenging work, because it contains two or more types of woven fabrics. A robust striped woven fabric pattern recognition method is presented in this paper, through which the striped woven fabric pattern could be segmented into three basic weaves and fancy woven fabrics based on Gray-Level Co-occurrence Matrix (GLCM). Firstly, scanning window is selected automatically by analyzing the characteristics of the striped woven fabric, and features are extracted in this window based on GLCM. Then we compute the correlation coefficient between the adjacent windows and complete the segmentation of striped woven fabric. At last, the segmented woven fabric patterns are recognized based on the approach of gradient histogram. According to the tests, we concluded that this method can segment and recognize the striped woven fabric patterns successfully, which can overcome the effects of thickness and color of yarns changing, and uneven illumination.

ESPI filtering method based on anisotropic coherence diffusion and Perona-Malik diffusion

Noise reduction is one of the most important concerns in electronic speckle pattern interferometry (ESPI). According to partial differential equation (PDE) filtering theory, we present an anisotropic PDE noise-reduction model based on fringe structure information for interferometric fringe patterns. This model is based on coherence diffusion and Perona-Malik (P-M) diffusion. The former can protect the structure information offringe pattern, while the latter can effectively filter off the noise inside the fringes. The proposed model generated by the two diffusion methods helps to obtain good effects of denoising and fidelity. ESPI fringes and the phase pattern are tested. Experimental results validate the performance of the proposed filtering model.

Driver Fatigue Detection Based on Eye State Recognition

Driving fatigue is a main factor caused the traffic accidents. Our faces contain a lot of useful information, we can use the state of eyes to detect the fatigue, but the eye state would be affected by wearing sunglasses. In this paper, to solve above problems and make the algorithm keep the accuracy and real-time at the same time, we use the infrared videos for detecting and propose an eye state recognition method based on convolution neural network (CNN), eventually calculating percentage of eyelid closure over the pupil over time (PERCLOS), blink frequency to detect the fatigue. The experimental results show that the proposed method has high recognition accuracy of state of eyes when wearing glasses and can detect the fatigue effectively.

Retinal vessel segmentation based on adaptive difference of Gauss filter

Based on the difference of Gauss (DoG) filter, a new retinal vessel segmentation method is proposed in this paper. Firstly, contrast limited adaptive histogram equalization (CLAHE) is used to improve the contrast of the image and then anisotropic diffusion equation is applied to smooth the image for the central reflex of the vessel. Secondly, adaptive DoG (ADoG) with different scale factor σ is used to give the initial vessel segmentation result. Then, the refined vessel enhancement result is computed by the superposition of ADoG in twelve directions. At last, the non-vessel is removed based on the bimodality of histogram of the image after enhancement and smoothing. We evaluate experimental results on the public DRIVE and STARE datasets qualitatively and quantitatively, and demonstrate the performance of the proposed method.

A supervised correspondence method for statistical shape model building

The construction of statistical shape model (SSM) is an important research topic in medical imaging benefited from its robust and nature represent of anatomical structures. Place-march of corresponding landmarks is one of the major factors influencing 3D SSM quality. In this paper, we present a supervised correspondence method for fast building SSM, which includes two main steps, i.e., surface data alignment and landmarks specified based on surface parameterization. The framework is validated with statistical models of the liver constructed from contrast CT images. The experiment results demonstrate that the generated model is statistical and anatomically meaningful.