Research On Image Processing Of Biological Information Based On Partial Differential Equations

Image processing of biological information is a branch of the bio-engineering disciplines, including biological information processing technology, biotechnology, image processing and analysis, Biological image processing and analysis, which is one of the fastest growing disciplines in the field of bio-engineering direction. Image analysis of biological information is committed to the digital information extracted from biological images or biological image sequences, there is a wide range of applications in the field of life sciences. However, biological information images often have a certain amount of noise, for subsequent image segmentation, it needs enhance the input image edge and improve image quality as well. On one hand, the biological information image filtering requires not only to remove blurring and noises, but also to keep the details of the image, while the traditional filtering methods are difficult to deal with such issues. On the other hand, nowadays, there has not been one unified scheme biological information image segmentation due to its unique complexity and diversity, traditional threshold segmentation method and watershed mark transform method are not suitable for complex biological images, and they will lead to segmentation failure.In recent years, the theory of partial differential equations has been widely used in various areas of image processing and achieved a great deal of achievements because of its strict mathematical theory. For images of biological information, combined with the theory of partial differential equations, the research work of this thesis are as follows:1. The traditional partial differential equations denoising model for biological information image may produce isolated point and noise enhancement, to solve the above problem, an improved anisotropic diffusion equation denosing model is designed. Firstly it uses morphological reconstruction technique to process the regularization item of images and then use the two diffusion coefficients respectively to handle edge region and non-edge region of image differently. Furthermore, in order to reduce unnecessary diffusion time, the threshold parameter is recorded as a function of time. The results of experiments show that the new method is able to retain more edge details, and it has better signal-to-noise ratio than the traditional PDE denoising model. 2. A problem of the total variation model is that it is easy to be affected of the noise in the image processing, which may lead to miss some important details, to solve the problem, an adaptable total variational denoising and enhancement model of the biological information image is proposed. The adaptive smoothing term can smooth the edge area and the flat areas in different degree keep the details on edges. The analysis of the new model and the experimental results show that the proposed variational model has better performance than the traditiona PDE models, and it is an efficient pre-processing method for image, a watershed transformation segmentation method with markers can extract the object region of biological information image by using the new model.3. Biological information image edge of the target area is usually blur, in order to segment the biological information image quickly and accurately, a novel active contour model for image segmentation using distance regularized term is proposed in the article. The model uses two different regularization functions for the global fitting term and the local fitting term to improve the accuracy of segmentation. A new edge stopping function is proposed to protect the weak edges of image. A distance regularization term using a function with two minimums is designed to eliminate the re-initialization of the level set function in the process of iterations. And also, a big time step is used to increase the speed of the evolution. The experimental results show that the new segmentation model is not sensitive to the settings of parameters and it can protect the weak edges of the biological information image effectively, and it has a strong segmentation robustness.4. A combination of regional neighborhood local binary fitting image segmentation model is proposed to segment biological information image to deal with the gray scale inhomogeneity and brightness inconsistency of the image, the new model uses the global fitting term which adds pixel neighborhood impact items to divide the biological information image fastly. Meanwhile, in order to ensure accurate calculation of the level set method, the energy functional uses distance regularization term to eliminate the re-initialization of the level set evolution process. The experimental results show that the model is not easy to fall into the local minimum regions, it has fewer iterations of evolution and is not sensitive to the location of the initial contour.