Researches On The Microscopic Image Segmentation Of Harmful Algae Based On Variational Level Set

In recent years, red tides with serious harm appear more and more frequently in coastal waters, which destroy the ocean environment seriously and restricts the continuable development of the economy of coastal regions in China. Our country arranged researches on mechanism, early warning, forecasting, prevention control methods of red ride from different aspects, the automatic monitoring of red tide plays an important role in preventing the occur of the red tides and classifying the dominant species of red tide rapidly and effectively, moreover the results of microscopic image segmentation for algal blooms affect the automatic monitoring. The main job of the paper is to research the methods of microscopic image segmentation for algal blooms.Variation and level set methods are two important and effective mathematical tools of partial differential equation. Image segmentation models based on the two tools contain parametric active contour-Snake model of variation method and geodesic active contour model of level set method, the two models embody the advantages of image processing based partial differential equation. Refer to the parametric active contour-Snake model is an energy minimization model and geodesic active contour model is easy in dealing with the adaptability of topological changes, and there are many kinds of harmful algal in the China's coastal waters , which have the different biological morphological features, moreover, the microscopic image is noisy and blurry, the paper puts forward an ideal method of microscopic image segmentation of harmful algal, which is based on the variational level set method.The method of image segmentation based on variational level set contains the edge-based method which uses image's gradient information and is not suitable to noisy and blurry microscopic images, and region-based method which uses the image's regional information and has better optimal segmentation effects. This paper analyzes and introduces the two classic models in detail, which are M-S model and C-V model. The energy function of the M-S model describes the image's regional information and the active contour, and gets the image's contour, region and smooth image via optimizing this energy function only once, although the original image is blurry and noisy. Somebody proposes using the level set method for solving the optimization question, which is to build C-V model of the simple M-S model. The C-V model does not rely on the image gradient information, besides the initial position of the level set formulation can local in any image's place, but the C-V model has the disadvantage of the level set evolution needing reinitialized continually. Therefore , this paper proposes the variational level set method without re-initialization of the image segmentation referring to the Li ChunMing's method which adds the distance regularization term-penalizing term in the energy function. Based on the method, by considering and utilizing both the edge and regional information, the author obtains the new method of the microscopic image segmentation which is suitable for the most algal blooms——edge and region based variational level set method of microscopic image segmentation of harmful algae. Finally the paper proves the effectiveness of the method via the results of the experimentation.