Search

The traditional neighborhood rough set model has low noise immunity and computational efficiency for mixed data. This paper establishes a local variable-precision neighborhood rough set model with dynamic change of attributes based on matrix theory. Firstly, on the basis of the updating law of local diagonal matrix and transition matrix, the dynamic updating mechanism of approximation under the local variable-precision neighborhood rough set of mixed information system is constructed; secondly, according to this dynamic updating mechanism, a new dynamic algorithm of local variable-precision neighborhood rough set with attribute change is proposed; finally, through the experimental analysis, it can be seen that the proposed dynamic algorithm has a high computational efficiency and a good robustness. [ABSTRACT FROM AUTHOR]

In order to solve problems of traditional image denoising methods such as image blur, edge information loss and so on, an image denoising method based on improved simplified pulse coupled neural network was proposed according to characteristics of underground images including uneven luminosity and large noise. Selection of neurons joining strength β was improved, which made β depend on pixel gray value of image, so as to get better denoising effect. At the same time, selection of decay time constant αE of dynamic threshold was improved, which made αE depend on amplification coefficient vE of threshold output, so as to reduce number of parameters of simplified pulse coupled neural network model. The value of vE was selected through experiment. The experiment results show that the method removes salt and pepper noise of underground images more effectively and preserves details of image edge more completely than traditional median filtering and mean filtering.