Applied Mechanics and Materials Vols. 519-520

Abstract: This paper present a way of pulmonary interstitial pathology diagnosis of computer-aided diagnosis based on wavelet analysis. It is difficult to diagnosis qualitatively in pulmonary interstitial pathology because various lesions of the image analogous and the image interlap. The method based on good time-frequency characters is put. The effectiveness and accuracy of the means is verified through the simulation experiment of denoised image, image segmentation and image characteristics extraction. Along with the further research and application of the wavelet technique, it will have more space that use wavelet analysis in computer-aided diagnosis.

Abstract: The local self-similarity descriptor is a kind of important image or video local feature description method. It is often used for detection, identification and recognition. In this paper we propose a new local self-similarity descriptor based on structural similarity (SSIM) index. It is showed in this paper that the SSIM Index give very different answers to the question of how self-similar local patches really are. For a given image we compute SSIM index distances between representations for all pairs of spatial-patches and store the results in a Self-Similarity Matrix (SSM) defined as the local feature descriptor. This new method is easily extended to the wavelet representation of images. Comparative evaluation of local feature descriptor with previous methods demonstrates improved performance.

Abstract: Making use of the three-dimensional reconstruction that poles features do to the two elements of the space and the quadric curve,we can make the polar lines of the image pairs.The polar lines in one of the features points meet and form the pole.According to the speciality the pole has in the optical center line of the CCD-camera and making use of the invariance of the rate of change of the theorem of the sequence in the projective transformation,we can find the corresponding matching points.According to the characteristics of their orders,their intersections form space objects.At last,accordiong to it,we can get the correct parameters of the space objects.

Abstract: This paper presents a method to extract leaf boundary from its 2D image. To solve this problem we firstly perform boundary feature extraction, based on which the key points on the boundary can be sampled and accordingly positioned in their geometry coordinates. Secondly, the leafs boundary configuration may be modeled with a wind noise method. Its amplitude development allows a linear sine function behaviors. The corresponding sampled points amplitudes in this model can be calculated according to their actual coordinates. Thirdly, to achieve better results to extract the discussed leaf boundary from its initial image, the model has to be self-optimized and self-improved ability by using a genetic algorithm.

Abstract: With the aid of TI CCS4.0 software development platform, using C language, a kind of track detection algorithm based on gray image is implemented on DM642. The detailed design and implementation steps are introduced. First, image median filtering, vertical sobel edge enhancement, and using the threshold segmentation method for binarization processing. Then search and mark connected domain on the basis of the binary images and prepare for extraction the connected domain about rail. Finally extract and fill around tracks to get a image containng only track region. The experimental results show that the design is convenient and feasible, and used in actual projects.

Abstract: The fisheye lens is a kind of ultra wide angle lens, which can produce a big super-wide-angle lens distortion. In order to cover a large scope of light, barrel distortion is artificially added to the optical system. However, in some cases this distortion is not allowed, then it requires calibrations of those distortions. Most of the traditional distortion calibration method uses target plane calibration to do it. This paper discusses the way of design fisheye lens, through which we can know the forming process of distortion clearly. Based on this paper, a simple and effective calibration method can be understood. Different from common camera calibration method, the proposed calibration method can avoid the error occurring in the process of calibrating test, that directly use the lens’ characteristic curve. Through multiple sets of experimental verifications, this method is effective and feasible.

Abstract: In this paper, a real-time video stabilization algorithm based on smoothing feature trajectories is proposed. For each input frame, our approach generates multiple feature trajectories by performing inter-frame template match and optical flow. A Kalman filter is then performed to smooth these feature trajectories. Finally, at the stage of image composition, the motion consistency of the feature trajectory is considered for achieving a visually plausible stabilized video. The proposed method can offer real-time video stabilization and its removed the delays for caching coming images. Experiments show that our approach can offer real-time stabilizing for videos with various complicated scenes.

Abstract: Age classification based on facial images is attracting wide attention with its broad application to human-computer interaction (HCI). Since human senescence is a tremendously complex process, age classification is still a highly challenging issue. In our study, Local Directional Pattern (LDP) and Gabor wavelet transform were used to extract global and local facial features, respectively, that were fused based on information fusion theory. The Principal Component Analysis (PCA) method was used for dimensionality reduction of the fused features, to obtain a lower-dimensional age characteristic vector. A Support Vector Machine (SVM) multi-class classifier with Error Correcting Output Codes (ECOC) was proposed in the paper. This was aimed at multi-class classification problems, such as age classification. Experiments on a public FG-NET age database proved the efficiency of our method.

Abstract: GPU based sparse reconstruction shows great significance in cone beam computed tomography (CBCT). This paper proposes a GPU based efficient algorithm for sparse view CBCT reconstruction. The reconstruction problem is converted to a constrained optimization using total variation minimization. The alternating direction method is adopted to solve it efficiently. Furthermore, a linearized proximity and FFT techniques are used for improving computation efficiency. To tackle with the most time consumption of forward and backward projection operation, the GPU hardware acceleration is utilized. The simulation experiments indicate that the new method is able to realize high accuracy reconstruction for CBCT with high speed.

Abstract: In this paper, we propose a coarse-fine matching algorithm based on parallax waveform analysis and conjoint measurement, which aims at performing the binocular stereo matching in natural scene. For the reason that human only notices the most saliency parts in image, we first extract the saliency areas of the two images, which suppresses the interference of background and reduces the computation cost. According to the saliency map, a global algorithm based on parallax waveform analysis is proposed to achieve the best coarse matching. On this basis, the fine matching is completed by analyzing the local features with the conjoint measurement of SAD and SSIM, which further improves the accuracy in binocular stereo matching. The performance of our algorithm is well demonstrated by experiments.