Research On Segmentation Of Complex Algae Cell Image

Algae is not only important aquatic resource, but also the perpetrators of red tide, eutrophication of water bodies and other natural disasters. The traditional manual statistical method requires heavy workload, and is also time-consuming as well as easy to make mistake .Therefore, to found simple, fast, accurate and convenient algae cell analysis, statistical techniques which also Consistent with the national standards, is of great significance to improve the detection of algae in water capacity and rapid monitoring of the aquatic environment. The digital microscope image method based on microscope images of algal cells and computer technique has unique advantage in algae cell automatic identification and analysis. However, as algae diversified in different kinds, algae cells have many special structural features in color, shape and texture, it is difficult to obtain effective segmentation results.In this paper, we focused on the image segmentation research of different algae in Chlorophyta, Bacillariophyta and Cyanophyta, which are the common dominant algae in freshwater lake. After deeply analysed these algae cells in shape, color, texture and other characteristics, several novel segmentation algorithms are proposed as follows:(1) The algae cell contour extraction algorithm based on Grab Cut and the 8-direction chain code edge tracking method. Firstly, the Grab Cut algorithm is applied to the filtered image. Afterward, we use the 8-direction chain code edge tracking algorithm to track the edge,At last, Fourier descriptors are applied for boundary smooth. The algorithm is especially applicable to the segmentation of algae cell image with complex background. For example, it is particularly effective in extracting the cell contour when the samples collected not only contain the algae cells, but also a large number of dirt and impurities.(2) The segmentation algorithm Based on color gradient method. As most of the algae cells have many different colors, we extract color gradient vector form the RGB color space, therefore it reserves and takes full use of color information and distinguish the algal cells with background effectively .Meanwhile, though boundary tracking and filling the inner holes, most cell images of algae in Chlorophyta and Cyanophyta can be effectively segmented. This algorithm is proved to have strong adaptability.(3)The segmentation algorithm based on the maximum entropy of the mean value-gradient co-occurrence matrix, which combined the color information with the maximum entropy. This algorithm built a two-dimensional histogram, and considered anti-noise performance of the mean value and the edge sharpening performance of gradient value comprehensively. As a result, the interference of the non-uniform background and the noise were avoided. This method is mainly applied to the Nitzschia algae in Bacillariophyta.