Yankui Sun

Automated Thickness Measurements of Pearl from Optical Coherence Tomography Images

In this paper, we explored the automatic thickness measurements of pearl from optical coherence tomography (OCT) images. We used a two stage scheme to extract the upper and lower boundaries of nacre respectively, and computed the thickness of nacre based on the extracted upper and lower boundaries. At the first stage, we employed edge detection method to extract the upper boundary. At the following stage, we used pixel classification method to detect the lower boundary. In both stages, boundary refinement and fitting were conducted. The proposed approach is evaluated using pearl optical coherence tomography images, and achieved high segmentation accuracy of 93.56% and relative measurement error of 1.69%. Experimental results demonstrate the effectiveness and robustness of our method.

A New Wavelet-Based Adaptive Algorithm for MR Image Enhancement

In this paper, a new wavelet-based adaptive enhancement algorithm for MRI has been proposed. In the new algorithm, two non-linear adaptive rules have been used to enhance both of lower and higher frequency wavelets coefficients separately. The experiment results showed that enhanced images obtained by new algorithm have better contrast of gray levels and less noise comparing with results from other wavelet-based adaptive enhancement algorithms.

Intelligent Diagnosis Method for Rotating Machinery Using Dictionary Learning and Singular Value Decomposition

Rotating machinery is widely used in industrial applications. With the trend towards more precise and more critical operating conditions, mechanical failures may easily occur. Condition monitoring and fault diagnosis (CMFD) technology is an effective tool to enhance the reliability and security of rotating machinery. In this paper, an intelligent fault diagnosis method based on dictionary learning and singular value decomposition (SVD) is proposed. First, the dictionary learning scheme is capable of generating an adaptive dictionary whose atoms reveal the underlying structure of raw signals. Essentially, dictionary learning is employed as an adaptive feature extraction method regardless of any prior knowledge. Second, the singular value sequence of learned dictionary matrix is served to extract feature vector. Generally, since the vector is of high dimensionality, a simple and practical principal component analysis (PCA) is applied to reduce dimensionality. Finally, the K-nearest neighbor (KNN) algorithm is adopted for identification and classification of fault patterns automatically. Two experimental case studies are investigated to corroborate the effectiveness of the proposed method in intelligent diagnosis of rotating machinery faults. The comparison analysis validates that the dictionary learning-based matrix construction approach outperforms the mode decomposition-based methods in terms of capacity and adaptability for feature extraction.

Fully automated macular pathology detection in retina optical coherence tomography images using sparse coding and dictionary learning

Abstract. We propose a framework for automated detection of dry age-related macular degeneration (AMD) and diabetic macular edema (DME) from retina optical coherence tomography (OCT) images, based on sparse coding and dictionary learning. The study aims to improve the classification performance of state-of-the-art methods. First, our method presents a general approach to automatically align and crop retina regions; then it obtains global representations of images by using sparse coding and a spatial pyramid; finally, a multiclass linear support vector machine classifier is employed for classification. We apply two datasets for validating our algorithm: Duke spectral domain OCT (SD-OCT) dataset, consisting of volumetric scans acquired from 45 subjects—15 normal subjects, 15 AMD patients, and 15 DME patients; and clinical SD-OCT dataset, consisting of 678 OCT retina scans acquired from clinics in Beijing—168, 297, and 213 OCT images for AMD, DME, and normal retinas, respectively. For the former dataset, our classifier correctly identifies 100%, 100%, and 93.33% of the volumes with DME, AMD, and normal subjects, respectively, and thus performs much better than the conventional method; for the latter dataset, our classifier leads to a correct classification rate of 99.67%, 99.67%, and 100.00% for DME, AMD, and normal images, respectively.

Medical Image Fusion by Multi-resolution Analysis of Wavelets Transform

A novel algorithm for the multimodalities medical images fusion based on wavelet transform has been proposed and implemented. The autoadaptive weighted coefficients have been calculated recursively to maximize the mutual information between the source image and the result image. Adopting multi-resolution analysis of wavelet transform, we achieved the MRI and CT image fusion. In addition, the new algorithm has been extended to MRI and color image fusion. The experiment results demonstrate that the new algorithm with wavelet transform have better fusion results compared with other mutual information fusion schemes without wavelet transform.

Method for optical coherence tomography image classification using local features and earth mover's distance

Optical coherence tomography (OCT) is a recent imaging method that allows high-resolution, cross-sectional imaging through tissues and materials. Over the past 18 years, OCT has been successfully used in disease diagnosis, biomedical research, material evaluation, and many other domains. As OCT is a recent imaging method, until now surgeons have limited experience using it. In addition, the number of images obtained from the imaging device is too large, so we need an automated method to analyze them. We propose a novel method for automated classification of OCT images based on local features and earth movers distance (EMD). We evaluated our algorithm using an OCT image set which contains two kinds of skin images, normal skin and nevus flammeus. Experimental results demonstrate the effectiveness of our method, which achieved classification accuracy of 0.97 for an EMD+KNN scheme and 0.99 for an EMD+SVM (support vector machine) scheme, much higher than the previous method. Our approach is especially suitable for nonhomogeneous images and could be applied to a wide range of OCT images.

Image denoising based on complex contourlet transform

In this paper, we propose a new image demising method based on complex contourlet transform. The complex contourlet effectively incorporate the approximate shift invariant property of dual-tree complex transform and directionality and anisotropy of contourlet transform. We apply our methods to image denoising experiments on gray and color images, and compare it with other popular denoising schemes. Our experiments show that the proposed method outperforms the adaptive Wiener filter, the wavelet shrinkage denoising and the contourlet denoising both visually and in terms of the PSNR values, especially for images with high texture information.

A two-dimensional lifting scheme of integer wavelet transform for lossless image compression

For the separable two-dimensional wavelet transform, its 2D lifting scheme is implemented by a sequence of one-dimensional wavelet transforms, and the corresponding 2D integer version is built in a similar way. In this paper, by introducing and investigating a matrix representation of the separable two-dimensional wavelet transform, we propose a new 2D lifting scheme and its integer version, where the lifting steps operate on the image directly. It is proved that they have less multiplication number than the classical ones, and the 2D integer wavelet transform is different from the classical 2D integer wavelet transform. Experiments indicate that, for the popular (5-3) filter, the 2D integer wavelet transform can achieves a little better compression than the classical one. The paper also compares JPEG2000 with JPEG-LS by experiments on DEM lossless image compression.

Automated thickness measurements of nacre from optical coherence tomography using polar transform and probability density projection

This paper presents a novel method for automated thickness measurements of nacre from optical coherence tomography (OCT) using polar transform and probability density projection. Our method detects the upper boundary of nacre and the raw boundary is fitted to a circle by a priori. Then polar transform is applied to the image according to the fitted circle. At the following stage, we extract the lower boundary by a transform incorporating unilateral difference and intensity of pixels. Finally, we get the probability distribution of nacre thickness by probability density projection. The proposed approach was evaluated using a large number of pearl optical coherence tomography images and achieved high accuracy, which could meet the requirements of practical applications. Besides, our method is robust to noise, boundary discontinuity, and it is real time. It has potential to be used in processing of some other OCT images with circle-like boundary.

SYMMETRIC LIFTING FACTORIZATION AND MATRIX REPRESENTATION OF BIORTHOGONAL WAVELET TRANSFORMS

This paper introduces some properties of symmetric Laurent polynomials, and then extends Euclidean algorithm to symmetric Laurent polynomials. The new results are used to investigate factorization of polyphase matrix for biorthogonal finite filters. It is shown that there exists one and only one symmetric factorization of the polyphase matrix, and the symmetric factorization can be determined directly and efficiently by Euclidean algorithm for symmetric Laurent polynomials. Finally, symmetric implementation and matrix representation of biorthogonal wavelet transforms are introduced, and the study demonstrates that the symmetric implementation has the least multiplication number in all lifting implementations, and it is equivalent to a matrix transform on finite dimensional vector space.