An efficient point pattern matching algorithm for articulated and multiple objects is presented in this paper. A local to global strategy is adopted to get three layered matches starting from an initial partition of t...
详细信息
ISBN:
(纸本)0769521282
An efficient point pattern matching algorithm for articulated and multiple objects is presented in this paper. A local to global strategy is adopted to get three layered matches starting from an initial partition of the point set using clustering method. Firstly, initial match with three point correspondences is obtained through local neighborhood and exhaustive search. Then, central match is expanded from initial match by alternating between next point matching and transformation update. Finally, ambiguous boundary points are matched and classified into their correspondent parts using the estimated alignment transformations, and missing or extra points are detected and rejected as outliers. Experiments on real images present satisfying results.
Guided by a building concept model, which interprets building into different levels and scales, this paper presents a method to extract buildings in monocular urban aerial images without priori illuminating or orienta...
详细信息
ISBN:
(纸本)0780381939
Guided by a building concept model, which interprets building into different levels and scales, this paper presents a method to extract buildings in monocular urban aerial images without priori illuminating or orientation knowledge. By using a shadow context model, a method was proposed to estimate the direction of shadow cast to verify the raw segmentations. Building extractions are refined by context, and a method of partial snake with the aid of the shadow cast direction is proposed, which can sharply reduce the iteration complexity and the influence caused by illumination. The extraction of self-shadow on gable roof with a proposed mathematical roof model is also discussed in this paper.
Multisensor image registration is a difficult problem. In this paper, we give a new registration method using direct histogram specification technique. We find that after using histogram specification, the resulting i...
详细信息
Multisensor image registration is a difficult problem. In this paper, we give a new registration method using direct histogram specification technique. We find that after using histogram specification, the resulting images with the same view look more similar, though the original images gained by different sensors differ much in intensity. Based on this property, a novel approach to find matching block pairs is proposed. The centers of the block pairs are used as control points (cps). We also use the cluster method of the nearest function criterion to test the correctness of the cps and discard wrong ones. The algorithm has been tested by many aerial images of different sensors. The effectiveness is illustrated by the experimental results.
This work presents a face detection method based on kernel Fisher discriminant analysis (KFD). Kernel based methods have been extensively investigated both in theories and applications, such as SVM and kernel PCA. Usi...
详细信息
ISBN:
(纸本)0769521223
This work presents a face detection method based on kernel Fisher discriminant analysis (KFD). Kernel based methods have been extensively investigated both in theories and applications, such as SVM and kernel PCA. Using the kernel trick, linear Fisher discriminant can be extended to non-linear case. Since the distribution of face patterns is very complex and highly nonlinear, using non-linear classification tools can hopefully tackle the problem of face detection. We explore the application of KFD in the task of frontal face detection. The experimental results prove the effectiveness of KFD in the face detection problem.
Text that appears in a scene or graphically added to video can provide an important supplemental source of index information as well as clues for decoding the video's structure and for classification, and we call ...
详细信息
ISBN:
(纸本)0780384032
Text that appears in a scene or graphically added to video can provide an important supplemental source of index information as well as clues for decoding the video's structure and for classification, and we call them closed caption. In this work, a novel algorithm is presented for detecting and locating caption in digital video. The first module of the system divides an image into small blocks featured by pixel value that is fed to SVM (support vector machine) to classify whether they are text blocks or not. The other module is to do post-processing on the classified text blocks to identify the rectangle region of them and OCR can be used further and easily. Experiments conducted with a variety of video sources show that our method could detect and locate caption region successfully by SVM with comparatively less samples.
This paper discusses a new algorithm of sub-pixels image matching and analyzes the characteristics of resampling and surface fitting methods. In order to meet the matching demands and to alleviate the computation work...
详细信息
This paper discusses a new algorithm of sub-pixels image matching and analyzes the characteristics of resampling and surface fitting methods. In order to meet the matching demands and to alleviate the computation workload, the following improvement algorithms are used. First, resample the model n-times, putt out (2n-1) sub-models, and calculate the NCs between each sub-model and image. Then choose the maximum between the sub-model and the displacement corresponding to this sub-model which requires the sub-pixel displacement. Finally, put forward a new algorithm that combines the resampling with surface fitting methods. Experimental results show the validity of the algorithm.
This paper proposes a novel approach for image lossless compression based on fuzzy logic and adaptive prediction. By a flexible strategy, the method can acquire a set of original predictors describing the more detail ...
详细信息
This paper proposes a novel approach for image lossless compression based on fuzzy logic and adaptive prediction. By a flexible strategy, the method can acquire a set of original predictors describing the more detail characteristic. Using a neural network, the proposed method can more efficiently organize the training of original predictors and implement adaptive prediction in fuzzy style. In entropy coding phase, the context-based conditional adaptive arithmetic encoding is adopted. The experiments demonstrate the characteristics make the approach achieve good tradeoff between computational complexity and efficiency of prediction and good performance for lossless compression.
A segmentation model that combines the Mumford-Shah (M-S) model and narrow band scheme of level set is presented. The M-S model is a desirable model for image segmentation, but computationally time-consuming. This pap...
详细信息
A segmentation model that combines the Mumford-Shah (M-S) model and narrow band scheme of level set is presented. The M-S model is a desirable model for image segmentation, but computationally time-consuming. This paper introduces a fast segmentation model, which combines the M-S model and narrow band scheme using new initialization method. The new initialization method is based on fast marching method, and the computing time is O(n). In each iteration step, the new segmentation model only deals with the data in narrow band instead of the whole image. Comparing M-S model and new narrow band M-S, experiments show that the two models can obtain almost the same segmentation result, but the computing time of new narrow band M-S model is much less than M-S model.
We present a method for personal authentication bsed on deformable matching of hand appearance. Authentication systems are already employed in komains that require some sort of user verification. In this work, active ...
详细信息
This paper proposes a new tracking mechanism for semi-automatic video object segmentation. An interactive video object segmentation tool is presented for the user to easily define the desired video objects in the firs...
详细信息
暂无评论