Research Of Image Segmentation Based On PCNN

At present, Currently, the application of the digital image processing in the military realm is more and more extensive, the image segmentation is an very important part of the digital image processing, accurate image segmentation and edge detection is the primary foundation of military target identification, the method of image segmentation is numerous, and the new method is continuously upspring.The main research object of this paper is Pulse coupled neural networks, so we introduced the model,working principle,application background and meaning of PCNN. So we aim at the characteristic of PCNN network, and use it in the image segmentation, but the effect of traditional PCNN segmentation method isn't very ideal. So in this paper,we make some improvements based on the traditional PCNN segmentation method, we use linear descending function to replace the exponential descending function which can lowered the complications of the method; and we also adopt the concept of entropy which can use maximize entropy as the principle of abeyance, so it can change the iterative blindness of abeyance, and after the improvements of the other parameters, we can get a better segmentation result.Based on this improved segment method, this paper will applied it in the color image processing,through the comparison of some common color space this paper chooses the NRGBspace. In the NRGBspace each image of a color image will be disposed respectively by PCNN and then combine the result of each paper by the textual method, then we can get the final segment result. After some simulations, we still can prove that the segment method based on PCNN can acquire a better result in NRGB space than in other color space, no matter in the outline or in details.At last, this paper use the PCNN in the interactive segment method. The traditional Live-wire is not good at weak edge detection, it is easy to emerge some mistakes; and the complications in the weight path search is also higher. To dispose these shortcomings this paper use the improved Canny operator to replace the Laplace operator which can ameliorate the weight functions, and use the autowave characteristic of PCNN as the short path search method which can lower the computation and change the shortcomings of weak edge detection, eventually we can acquire better segmentation results.