Applied Mechanics and Materials Vols. 556-562

Abstract: Night-vision devices are becoming more and more important now, but images captured by these devices often have unpleasing visual effects. According to the characteristics of night-vision images, in this paper we propose an image enhancement approach maintaining color fidelity. Firstly, the image is divided into RGB channel and decomposed into reflectance component and illuminance component. Secondly, the non-uniform illuminance component of the image is filtered again to eliminate the uneven illumination. Finally, we combine the reflectance component and the corrected illuminance component and adjust the brightness of the combined image. The experiment result shows that the proposed enhancement approach has high-performance.

Abstract: It is well known that adaptive Gaussian chirplet signal decomposition algorithm has the best time frequency resolution in all signal decomposition algorithms. It is widely used in non linear and non stationary signal decomposition, especially for the signal which is superposition of chirplet functions decomposition. But it has a large amount of computation. In this paper, we propose a fast algorithm based on short time Fourier transform (STFT) method and we change parameters’ domain. Using this fast algorithm to decompose a four atoms non-linear signal computes very fast and it can also avoid the cross term’s interferer of the Wigner-Ville distribution.

Abstract: As a new image copyright protection technique, a novel image fingerprinting algorithm based on improved image normalization and DFT is proposed in this paper. We first research the DFT amplitude coefficients deeply and improve the traditional image normalization method to fit for more general images, including aspect ratio and rotated angle estimation and correction schemes before normalizing the image. Then combine the improved image normalization method and DFT together to extract the image fingerprint according to the DFT amplitude coefficients. Compared with others’ achievements, our method performs better in resisting common image processing attacks and geometric distortions, also some kinds of combined distortions.

Abstract: In this paper, a novel image classification method, incorporating active learning and semi-supervised learning (SSL), is proposed. In this method, two classifiers are needed where one is trained by labeled data and some unlabeled data, while the other one is trained only by labeled data. Specifically, in each round, two classifiers iterate to select useful examples in contention for user query. Then we compute the label changing rate for every unlabeled example in each classifier. Those examples in which the label changing rate is zero and the label in the two classifiers is the same are selected to add into the training data of the first classifier. Our experimental results show that our method significantly reduced the need of labeled examples, while at the same time reducing classification error compared with widely used image classification algorithms.

Abstract: This paper addresses semantic image classification with topic model, which focusing on discovering a hidden semantic to solve the semantic gap between low-level visual feature and high-level feature. In our approach, Latent Dirichlet Allocation (LDA) model successfully reflect the high level features and the RGB SIFT features which integrating the Scale-invariant feature transform (SIFT) features with color features on the assumption that pictures generated by mixture of latent semantic which we called topics. The proposed approach has a sufficient theoretical basis and the experimental evaluations the COREL database demonstrate its promise of the effectiveness.

Abstract: Fungal diseases are the major diseases of agricultural production, and have brought tremendous impact to it. Identification of spore morphology plays an important role in the identification of fungi. This paper uses the microscopy images of two kinds of fungal spore and utilizes the technology of image analysis and recognition to classify them. We firstly get the underlying feature descriptors of these two kinds of microscopy images by RGB SIFT (Scale Invariant Feature Transform), then create the visual word dictionary using K-means clustering algorithm, at finally we use LDA, KNN and SVM to classify these two kinds of images. The results indicate that the classification of spore image based on feature extraction is feasible. In our future work, we will conduct the classification of related species and highly similar spore images.

Abstract: How to get good time-frequency distribution of FH signals is crucial to detect and to trace the FH signals. The theory about the rearrangement of time-frequency distribution is summarized in the paper, then the basic principle of detecting the signals using the algorithm of rearrangement of the smooth pseudo Wigner-Ville distribution (RSPWVD) is analyzed in detail, and the algorithm expression are given at the same time. The analysis results show that the RSPWVD algorithm not only has more ideal anti-jamming effects, but also can enhance the time-frequency aggregation of the signal, so the signal parameters are estimated more accurately. In the end, the computer simulation is shown to test the feasibility and effectiveness of the algorithm.

Abstract: Tongue inspection is the main content of four diagnostic methods “observation, auscultation and olfaction, interrogation, palpation” for traditional Chinese medical science. Traditional tongue inspection mainly depends on doctor’s eyesight and subjective observation to diagnose a disease. It lacks objectivity and hinders further development. It is the only tendency to use computer technology to objective and standard tongue inspection. The tongue split out from the face is the premise for tongue objective. This paper introduces a new dual-threshold segmentation algorithm for tongue image base on HIS.

Abstract: Recently bag-of-words (BoW) model as image feature has been widely used in content-based image retrieval. Most of existing approaches of creating BoW ignore the spatial context information. In order to better describe the image content, the BoW with spatial context information is created in this paper. Firstly, image’s regions of interest are detected and the focus of attention shift is produced through visual attention model. The color and SIFT features are extracted from the region of interest and BoW is created through cluster analysis method. Secondly, the spatial context information among objects in an image is generated by using the spatial coding method based on the focus of attention shift. Then the image is represented as the model of BoW with spatial context. Finally, the model of spatial context BoW is applied into image retrieval to evaluate the performance of the proposed method. Experimental results show the proposed method can effectively improve the accuracy of the image retrieval.

Abstract: There are many difficulties in image overlapping in 3D terrain of dynamic LoD quad tree model, it will cause gaps between terrain tiles and images if both of them are expressed independently. This paper breaks through the LoD mesh difficulties, and realizes image overlapping based on GPU programming without modeling. It establishes different texture coordinate transformational relationships in GPU to realize seamless expression between terrain texture and overlapping image based on setting terrain vertexes with global unified texture coordinate. The affine transformation is adopted to solve the difficulty of multiple spatial nonlinear transformations aiming to the interaction of image overlapping, and also to solve the non-synchronous problem of texture coordinate in interaction. At last, the conclusion may be achieved. Image overlapping technique in LoD terrain without modeling based on GPU is a typical case of render technology to solve application difficulties, the realization is simple, efficient and steady which only faces to single terrain tile. This research will greatly enhance the ability of Digital Earth, which completely clear an important technical barriers and promote the applications in GeoDesign.