Entity

Time filter

Source Type


Li F.,Jiangsu Science Food College
Journal of Convergence Information Technology | Year: 2012

In order to segment cells image better, this paper modifies traditional BP neural network: first, set afferent neuron as a 3×3 window to replace traditional one pixel channel; second, apply a method based on comentropy to estimate the number of hidden neurons; at last, apply an improved PSO algorithm to guarantee that net weight values are converged to the optima. The segmentation results on blood cells and intestinal cells show that the number of hidden neurons this paper get greatly decrease with the false rate superior to that of BP algorithm and LM algorithm and the elapsed time smaller than that of BP algorithm and approaching LM algorithm. The method of estimating hidden neurons this paper put forward is effective, and improved PSO algorithm can escape from local minima and converge to global minima. Source


Fengling L.,Jiangsu Science Food College
Journal of Convergence Information Technology | Year: 2012

To get better filtering effect, this paper proposes a new adaptive weighted median filter algorithm for the weaknesses in the traditional median filter algorithm. this algorithm first identifies the noise points in the image via the noise detection, next adjust the size of the filter window in adaptive manner by the number of the noise points in the window, group pixel points in the filter window by the specific rule in an adaptive manner, give the weights of the pixel points in different groups by similarity and finally filter the detected noises by using the median filter algorithm. the computer simulation test results indicate that this new algorithm can not only effectively eliminate noises, but also better save image detail. Its filtering performance is better than it of the traditional median filter algorithm. Source

Discover hidden collaborations