The Study Of Digital Image Processing Of Biological Information

This dissertation deals with the study of cell image segmentation and de-noising methods used in the 3-dimension reconstruction of slice cell image. Main content as following:First, this dissertation describes the state and the development of the 3D reconstruction of slice cell image. It points out the registration and segmentation method are critical for the 3D reconstruction of cell slice image, which is determined by the limitation of cell slice production and complex property of biological cell, including animal cell and plant cell.Second, it reviews the state and the development of modern and traditional image segmentation technology's application in cell slice image segmentation and summaries that it is difficult to generally segment any kind of biological cell slice image automatically because of cell's complex structure and cell slice image's non-even gray distribution. It points out that general automatic cell slice image segmentation will be only achieved if visual mathematics model corresponding to mammalian vision systems is setup entirely.Third, it has a discussion on the development of modern and traditional artificial neural network technology and their application in image processing. It summaries that the model of the Pulse-coupled neural network (PCNN) based on Elkhorn's model of the cat visual cortex was constructed according to the mammalian vision systems. The nonlinear modulation property, which is a generic and pervasive biological pulse-coupling mechanism, is used in the construction of PCNN. So, PCNN finds many applications in image processing, including segmentation, edge extraction et al. As all known, the performance of the image segmentation depends not only directly on the adjustment of PCNN parameters and the statistical properties of image but also on the cyclic iteration times N of PCNN. If the parameters have been properly set, it turns out to be essential to select a suitable criterion to determine N. While N is usually evaluated by means of visual judgment which decreases the efficient of PCNN image segmentation. Here the dissertation raises a new method to implement the image segmentation automatically based on the PCNN model and the entropy of image. That is, the criterion of maximal entropy of segmented binary image of PCNN output. According to this criterion, the iteration times, N, is determined automatically.Traditional image segmentation algorithms exhibit weak performance for plant cell, which have complex structure. On the other hand, pulse-coupled neural network should be suitable to the segmentation of plant cell image. But the present theories cannot explain the relationship between the parameters of PCNN mathematical model and the effect of segmentation. Usually satisfactory results require time-consuming selection of experimental parameters. To avoid this flaw, this dissertation uses the maximum entropy principle for automatically segmenting plant embryonic cell image and the algorithm produces desirable result. Therefore, a model with proper parameters can automatically determine the number of iteration of PCNN, avoids visual judgment, enhances the speed of segmentation. Then, the segmented cell image will be utilized subsequently by accurate quantitative analysis of micro-molecules of plant cell. This criterion of maximal entropy is valuable for theoretical investigation and application of PCNN.Forth, A new, simple method of counting and segmenting cell image is suggested in this dissertation. It is based on the feature of cell's geometric and morphological information. The segment effect is tested with blood cell image and the result is desirable. It is very useful for cell image where the cells don't contact closely. Particularly does this method can count the number of blood cells, however, this counting is not very accurate, but it is enough to biological research of cytological count. At the same time this method can segment special blood cell from its neighborhood, it is very important for the development of image segment technology, because traditional image segment method is inadequate to achieve such segment and count of cells.Fifth, Applying the main idea of the improved center weighted median filter to the design of the optimal stack filter based on omnidirectional structural elements constrains, the dissertation raises a kind of improved filter algorithm for de-noising of pulse noise. The theory analysis and the simulation experiment of the image processing shows that this kind of filter can not only remove noise effectively but also keep detail of the image sufficiently.Sixth, Using the ability of the better keeping the image's detail of the morphological filter with omnidirectional multiple structuring elements and the ability of the strong removing noise of the improved center weighted median filter, this dissertation suggests a new kind of designing project of the hybrid filter for de-noising of pulse noise, which applied the idea of the stack filter based on omnidirectional structural elements constrains to the project. The theory analysis and the simulation experiment of the image processing indicate that result is desirable.The intensity of pulse noisy pixels significantly differs from adjacent pixels of image. The de-noising will be accomplished by modifying the intensity of the noisy pixel according to the firing patterns of adjacent neurons of PCNN. At last, this dissertation suggests a new kind of filter for de-noising pulse noise based on the PCNN and median filter. It is very simple, and the simulation experiment of the image processing shows that the de-noising effect is much better than neighborhood averaging method and median filter.