At present the most widely used technology of pinyin-Chinese character conversion combines statistics with linguistic rules. Although it basically solves such problems as long distance restriction and language recursi...
详细信息
At present the most widely used technology of pinyin-Chinese character conversion combines statistics with linguistic rules. Although it basically solves such problems as long distance restriction and language recursion phenomenon,it relies on a great deal of computation because there are too many candidate paths. This paper tries to simplify the candidate paths by using quotient space granularity computation theory, first obtains the scope of the best path in the coarser granularity world, and then uses more language rules to obtain the best path. The experiments indicate that this method can reduce the computation, speed up the conversion, and enhance the rate of accuracy by nearly 2％.
A novel and efficient speckle noise reduction algorithm based on Bayesian wavelet shrinkage using cycle spinning is proposed. First, the sub-band decompositions of non-logarithmically transformed SAR images are shown....
详细信息
A novel and efficient speckle noise reduction algorithm based on Bayesian wavelet shrinkage using cycle spinning is proposed. First, the sub-band decompositions of non-logarithmically transformed SAR images are shown. Then, a Bayesian wavelet shrinkage factor is applied to the decomposed data to estimate noise-free wavelet coefficients. The method is based on the Mixture Gaussian Distributed （MGD） modeling of sub-band coefficients. Finally, multi-resolution wavelet coefficients are reconstructed by wavelet-threshold using cycle spinning. Experimental results show that the proposed despeclding algorithm is possible to achieve an excellent balance between suppresses speckle effectively and preserves as many image details and sharpness as possible. The new method indicated its higher performance than the other speckle noise reduction techniques and minimizing the effect of pseudo-Gibbs phenomena.
Barcode has been widely applied in the modern world. This paper presents a fast and robust recognition method of noisy code 39 barcode. The proposed method can be divided into two steps: search and decoding. In the fi...
详细信息
Barcode has been widely applied in the modern world. This paper presents a fast and robust recognition method of noisy code 39 barcode. The proposed method can be divided into two steps: search and decoding. In the first step, all asterisks in the image are found with evenly defined scan lines and then those with the same directions are matched together to get a valid barcode region. In the second step, a local denoise method is first applied to eliminate noise in the barcode region and then a middle band filter is used to decode the barcode. Our method is simple in comparison with former methods and experimental results show that it is efficient for fast barcode recognition on noisy images.
This paper aims to carry out granular analysis of time sequence based on quotient space. Granular methods have long before been adopted to analyze time sequence, but the granularity was based on time, for example, day...
详细信息
This paper aims to carry out granular analysis of time sequence based on quotient space. Granular methods have long before been adopted to analyze time sequence, but the granularity was based on time, for example, day mean, month mean, year mean and so on in finance forecast. In this paper, the granularity is based on space and some significant results are obtained: we can, in certain circumstances, get characteristics of time sequence in an original space when carrying out granular analysis of it in its coarser-grain space; granular analysis of a Markov chain is equivalent to an hidden Markov model (HMM), contrarily, any HMM is equivalent to granular analysis of a Markov chain. These results deepened our understanding of HMM from the perspective of granular analysis. We can not only use the methods of HMM to study time sequence, but also use the methods of granular analysis based on quotient space theory to solve the problems of HMM.
A novel and efficient speckle noise reduction algorithm based on Bayesian wavelet shrinkage using cycle spinning is proposed. First, the sub-band decompositions of non-logarithmically transformed SAR images are shown....
详细信息
A novel and efficient speckle noise reduction algorithm based on Bayesian wavelet shrinkage using cycle spinning is proposed. First, the sub-band decompositions of non-logarithmically transformed SAR images are shown. Then, a Bayesian wavelet shrinkage factor is applied to the decomposed data to estimate noise-free wavelet coefficients. The method is based on the Mixture Gaussian Distributed (MGD) modeling of sub-band coefficients. Finally, multi-resolution wavelet coefficients are reconstructed by wavelet-threshold using cycle spinning. Experimental results show that the proposed despeckling algorithm is possible to achieve an excellent balance between suppresses speckle effectively and preserves as many image details and sharpness as possible. The new method indicated its higher performance than the other speckle noise reduction techniques and minimizing the effect of pseudo-Gibbs phenomena.
Quotient space theory of problem solving, a formal model of granular computing, is generalized in the sense that topological structure is replaced by Cech's closure space. Some basic issues of granular computing, ...
详细信息
In this paper, we propose a dimension reduction method of locality preserving projections based on QR-decomposition of training data matrix, namely LPP/QR. It is efficient and effective in under-sampled recognition of...
详细信息
The theory of granule computing based on the quotient space is one of the three main granule computing theories. The emphasis is on the structure of the quotient space theory in this paper. Comparing with Rough Set th...
详细信息
In this paper, we propose a dimension reduction method of locality preserving projections based on QR-decomposition of training data matrix, namely LPP/QR. It is efficient and effective in under-sampled recognition of...
详细信息
In this paper, we propose a dimension reduction method of locality preserving projections based on QR-decomposition of training data matrix, namely LPP/QR. It is efficient and effective in under-sampled recognition of image and text data, especially when the number of dimension of data is greater than the number of training samples. Its theoretical foundation is presented. The equivalence between LPP/QR and generalized LPP is induced although LPP/QR is faster than generalized LPP. Several experiments are conducted on Yale face database. High recognition rates show that the algorithm performs better in under-sampled situations.
暂无评论