Research On Pulse Coupled Neural Network In The Ultrasonic Image

In recent years, with the rapid development of interdisciplinary research, peopledemand a higher medical image processing results. Many scholars have studied outsome related algorithms about how to process medical image. But the difficult and hotissue of medical imaging processing is to explore algorithm with high speed andprecision. Pulse Coupled Neural Network (PCNN) is a new artificial neural networkwhich came from the research of visual properties, namely, the synchronous pulse burstphenomenon of cat’s visual cortex. At present, the research of PCNN is still in its initialstage whether at home or at abroad. It needs to be expanded and deepened in the future.We reference to previous thought, analysis and compare to the PCNN modelexisting in the literature, choose the reasonable PCNN simplified model which wedecided to use in the thesis. Then introduce of its application on image processing. Inevery experiment step, the related application problem of PCNN model was resolvedby the study of each parameter in the model. Doing research on PCNN expands theunderstanding of its theory.In this paper, the following works have been done according to the latest researchin PCNN:(1) To the speciality of mixed noise constituted by pulse noise and Gauss noise,we present a comprehensive algorithm, which is based on the simplified PCNN model,and combined with mathematical morphology method, median filtering and wienerfiltering. Through the comparison of traditional noise reduction algorithm, thisalgorithm not only can effectively filter out mixed noise interference, but also it canprotect the image edges and detail information. The image after noise reduction has abetter subjective visual effect and objective evaluation index performance. Especiallyin the signal-to-noise ratio and noise reduction ability are improved to some extent(2) PCNN is introduced into the medical ultrasonic image processing field.According to the characteristics of ultrasonic image, we proposed an ultrasound imageenhancement algorithm based on PCNN and simulated the human visual characteristicby changing the PCNN network parameters. The computer simulation shows, when using the enhancement algorithm in this paper, we can get better experiment resultsthan using the traditional classical algorithm. The algorithm can provide ultrasonicimage with clear content for clinical. This result will be facilitate the doctor diagnosedobservation as well as be advantage to the further processing of ultrasonic image. Thismethod has significant meaning for the follow-up medical image automatic focussegmentation and has improved the accuracy of computer-aided diagnosis.(3) The focus segmentation algorithm to the ultrasonic image was provided in thepaper. Aim at the problem which is hard to determine the parameters for the PCNN inthe past segmentation algorithm, so the new image segmentation method was proposedthat banded automatic optimization ability of PSO and used the improved maximumentropy function as the fitness function. The termination conditions is the maximumentropy function gets its max value. This algorithm realized automatic segmentationbased on the particle swarm optimization (PSO) and the PCNN; it can set PCNNparameters automatically, save the artificial interaction and experiment trouble, andsegment the focus area of ultrasonic image in the output results.