Applied Mechanics and Materials Vols. 263-266

Abstract: This thesis tries to calculate quantity and distribution of heat source in vivo from body surface temperature. A superposition temperature image of a number of point heat sources is surface fitted to get the Q of heat sources (information of cancer cells) quantitatively. More precisely, surface fitting the surface of temperature distribution based on T formula of multiple heat sources with MTLAB software. The result can reflect the disease area information because cancer cell’s Q value is much higher than that of normal cell. This application is a new try in the diagnosis of breast cancer, has an important value on the early detection and diagnosis of disease source.

Abstract: Wear debris recognition,which is based on patch similarity of anisotropic diffusion algorithm and BP neural network,is researched.At first, feature parameter refining of wear debris image,which was based on patch similarity of anisotropic diffusion algorithm feature parameter refining,was researched.Second,wear debirs classifier,which was based on the BP neural network and the first step,was researched.At last,with experiment results and experiment results analysis,the wear debris recogniton system in the paper is proved to be some merits,which include high classification accuracy,fast learning convergence rate and high recognition rate.

Abstract: In allusion to non-exactness edge positioning in the process of place a three-dimensional model, a method of multi-scale wavelet transfer is applied to obtain edge signals from satellite image and perform pixel dot correction, which is proved to give prominence to the landmark edge curve without avoiding sacrificing the precision position of the target landmark edge .This method, with fixed dot correction, makes it more convenient and more precisely to settle down edge positioning in the process , in order to place a three-dimensional model.

Abstract: In the information storage, the image is often encountered, but the image storage and analysis are very complex, in order to more save memory space and more be used as a compressed representation of a graph, we give a new definition in the paper. We show some properties and give the lower integral sum number of some graph.

Abstract: One method is proposed to remove the random noise and low-frequency coherent noise in the images of the optic 4f system, which is based on fusion of multiple spatial frequency spectrum images. The multiple spatial frequency spectrum images of once experiment are captured based on the image copying character from the lattice structure of spatial light modulators, which contains the same useful image information and noise with the similar distribution but different values. The random noise and coherent noise in these images are removed by combining it utilizing image fusion technique, which is similar to cumulative mean in time domain. The theoretical analysis and physical experiment suggests that the method is able to remove random noise and low-frequency coherent noise effectively and reserve the useful image information.

Abstract: The automotive night vision system can significantly improve the safety of driving at night by fusion of visible and infrared images. Due to the aim at helping driving at night, it’s important to develop the visibility and the consistency of the image and the real condition on the road. In this paper, visible light image is enhanced its light and shadow information based on nonlinear exponential function and the infrared image is increased its contrast by S-Retriex method. Then the enhancement images are fused and the color space of fusion image is adjusted by the reference statistical characteristics. Simulations have shown that the result image has good visual scene which retains the details of the infrared image and has original light, shadow and color information.

Abstract: Based on synthesizing quaternion and the R, G, B component of color image , a color image edge detection algorithm was proposed combining quaternion with Echo State Networks(ESN) in this paper . The pixel and its eight neighbourhood average is used to calculate vector product as the characteristic vectors of image edge. Then the ESN has been trained by the characteristic vectors. After training , the ESN is directly used for edge detection. Experiments show that this method has good effect of fabric edge detection, and have a stronger ability to keep the details.

Abstract: To support content based image retrieval, MPEG-7 is developed to define the content interfaces for images. In MPEG-7, Dominant Color Descriptor (DCD) is considered as the most important feature, and is widely used to describe the color features of an image. To support semantic queries from users, we proposed a color feature semantic mapping method in this work, which can translate the DCD values into semantic color names. The semantic mapping method is realized by constructing a mapping table between the DCD values and the semantic color names. To validate the effectiveness of our mapping method, an image retrieval experiment is conducted. From the comparison with the manually indexed description, the proposed mapping method is proved to be effective by the experiment results. Our work is very important to automatically generate the semantic description of an image and then support the users’ semantic retrieval queries.

Abstract: According to the infrared image noise interference serious and low signal to noise ratio characteristics, an image mosaic algorithm based on pulse coupled neural network is presented in this paper. The algorithm uses the PCNN imitation of biological vision to extract edge information, adopts the Hausdorff distance as the registration similarity measure, afterwards, computes the optimal registration parameters, simultaneously, uses the weighted average fusion method to achieve image mosaic. The experimental results show that the algorithm can achieve precise image mosaic and have better robustness to noise.

Abstract: With the development of the hyperspectral remote sensing technology, the level of quantitative remote sensing has risen greatly, but varying degrees of random noise is contained. To optimize the use of the remote sensing images and to improve the effectiveness and accuracy of discriminating ground objects according to spectral absorption features, random noise removing is necessary. Aimed at the distribution characteristics of the random point noise of HJ_1A HSI data level 2 product, eliminate the Gaussian white noise by soft-thresholding filtering based on the Birge-Massart penalty function in wavelet domain. And the result shows that the method can not only remove the additive noise effectively, but also retain most of the feature information and edge details of the original image. Therefore, the method could provide necessary support for quantitative use of HJ_1A HSI data.