Search

An abstract of the article "An image retargeting algorithm based on strip dividing and resizing" by Hua Shun-gang and colleagues is presented.

A set of multiple images for a person provides much richer appearance information than a single image. Extracting the common information of multiple images can improve face recognition performance. The classic principal angle-based approach MSM does not work well while appling in single test image matching with image set case. A new approach based on the criterion of minimum reconstruction error from the principal components of test images is proposed. Some experiments are implemented on testing face image sets extracted from MBGC data set, and the experimental results show that the proposed approach possesses significant performance upgrade and is superior to the single image-based and principal angle-based approaches. [ABSTRACT FROM AUTHOR]

A hierarchical phase correlation palmprint recognition (HPCPR) algorithm is presented for the low location accuracy palmprint image recognition. The global image matching and alignment are accomplished using the peak value and peak position of phase correlation function, and then a fine matching is realized through computing the weighted phase correlation of all sub-blocks taken from the coincident area of the two images. As HPCPR algorithm employs both global and local features of palmprint image and can quickly align candidate matching images, it outperforms the conventional phase correlation method in both recognition rate and efficiency. Experiments on PolyU Palmprint Database verified the effectiveness of this approach. [ABSTRACT FROM AUTHOR]

To solve the drawback that typical mutual information-based registration neglects the spatial information of images, a new medical image registration method is developed by combining mutual information with gradient similarity. The gradient similarity of each pair of corresponding points includes direction similarity and module similarity. The summation of the similarity term for all sample pairs gives the gradient similarity of two registration images, which is multiplied by the mutual information to form the final registration metric. Registration functions are analyzed and compared, applying in different transform such as translation, rotation and sub-sampling of 2D multi-modal images, respectively. Experimental results show that the new method performs better than typical mutual information in robustness and precision. [ABSTRACT FROM AUTHOR]

An innovative 3D modeling method for the hip joint and reparative modeling for femoral head based on CT images and the reverse engineering is proposed. The processing scheme consists of the following steps: first, detecting bone tissues from CT images and segmenting the femoral head and acetabulum respectively by model based on segmentation method. Then, the contours of femur and acetabulum are recorded as the style of points. After being reduced, the part point data of femoral head and acetabulum are extracted respectively and fitted by the way of the least square. Accurate reparative results are achieved from two collapsed femoral head of ANFH patients by the proposed method. Experiments show that the new method can reconstruct hip joint model and regain femoral head of patients perfectly. It can be developed to other joints or bone tissues too. The method affords theoretical model for accurate operation position fixed in orthopedics clinic and biomechanics analyses, and it also provides more effective ways for the individually manufacturing of artificial hip joint. [ABSTRACT FROM AUTHOR]

Neighborhood preserving embedding (NPE) is a subspace learning algorithm,which has the property of preserving local neighborhood structure on the data manifold. Although NPE has been applied in many fields, it. has limitations to solve recognition task. To improve the recognition performance of NPE, a new method, called neighborhood preserving discriminant embedding (NPDE), is proposed for face recognition. NPDE effectively combines the ideas of LDA and NPE, i. e. it can hold the strong discriminating power while preserving the intrinsic geometry relations of the local neighborhoods according to prior class-label information. Experimental results on ORL face database and Yale face database demonstrate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Gabor face representation has been getting popular in face recognition applications. However, it also suffers from the high dimensional data containing diverse redundancy and different random noises. To utilize the Gabor feature for efficient face recognition, a new Gabor feature selection method is proposed. Firstly, the Gabor feature differences between every two face images within a training data set are calculated and grouped into two categories; intra-individual set and extra-individual set. Then the rank of discriminating capabilities of features can be estimated by evaluating the classification error on intra-set and extra-set based on weak classifier built by single feature. The Gabor features with small errors were selected. And at the same time, the mutual information between the candidate feature and the selected features was examined. As a result, the non-effective features carrying information already captured by the selected features will be excluded. The features thus selected are both accurate and non-redundant. Finally, the selected Gabor features were classified by PCA and LDA for final face recognition. The experiments on CAS-PEAL large-scale Chinese face database show that the proposed method can greatly reduce the dimensionality of Gabor features and effectively increase the recognition accuracy. [ABSTRACT FROM AUTHOR]

The leaf shape design of multi-leaf collimator used in 3-dimensional conformal radiation therapy was dealt with. According to the requirements of precision conformal dose delivery in radiation therapy, the focused trapezoid cross-sectional shape of the flank leaf of multi-leaf collimator is introduced, and the tongue-groove structure, which is used to minimize interleaf transmission, is determined. Genetic algorithm is utilized to optimize the leaf end shape to improve the size and uniformity of leaf-end penumbra. For several typical field sizes, the optimization is carried out. The result shows that the leaf-end penumbra keeps nearly constant in any leaf position. Optimum dose conformation can be realized by these leaf shape designs. [ABSTRACT FROM AUTHOR]

A novel fingerprint matching algorithm was developed. In the alignment stage, a hybrid alignment algorithm was proposed, which used local orientation field and local minutiae topologic structure to achieve the alignment of two fingerprints. In the aligned minutia matching stage, a minimum set of matched minutia pairs was achieved by matching two minutia set by the elastic matching algorithm and, with the minutiae in the matched pairs set as node and straight lines connecting two minutiae as the edges, two corresponding graphs for input fingerprint and template fingerprint were constructed, and a matching algorithm was presented to match the two graphs. The matching method using global orientation field was also introduced to improve the accuracy of two fingerprints matching. Experiments on four databases of FVC2002 show that this fingerprint matching technique is effective. [ABSTRACT FROM AUTHOR]

A method of fusing multiple photographs into a high dynamic range image is presented. Based on the assumption of reciprocity, the photographs of the same scene are taken with different amounts of exposure by digital camera. Using the least squares principle and B-spline function, the response curve of imaging process can be recovered, and the mapping relationship between pixel values and exposures is acquired. On the basis of the response curve, a high dynamic range image, in which the pixel values are proportional to the true radiance values in the scene, can be constructed from the multiple photographs. It will cover the entire dynamic range captured by the original photographs, and some details information of high light or black regions in the scene can be made Out. The experimental results show that the constructed high dynamic range images are satisfactory. [ABSTRACT FROM AUTHOR]

The research on the conversion from 2D paper drawings to 3D computer feature models has been stuck in low-level coding. One of the reasons for it is that the three phases of understanding process has been isolated and people have been doing their research on only one of the phases since the whole conversion is very complicated and more difficult. For instance, the vectorization methods for the first phase were developed only for getting straight lines, arcs, circles, etc. so that much information contained in the drawing has been lost after the vectorization. This paper develops a holo-extraction method of information from paper drawings, the networks of single closed regions (SCRs), which can not only provide a unified base for recognizing both the annotations and the outlines of projections of parts, but also build the holo-relationships among SCRs so that it is convenient to extract lexical, syntactic and semantic information in the subsequent phases for 3D reconstruction. [Copyright &y& Elsevier]