Segmentation Algorithm Based On Region Growing And Clustering Flowers Image

The typical definition of the image segmentation problem is extracted from theimage processing consistency during the process of image regions of interest, that is,the interesting part is extracted in order to analyze an image. Image segmentation isan important part of the digital image analysis and image processing, It is afundamental and critical technology. It plays an important role in study of image as atest of image preprocessing and a basis of image recognition, analysis andunderstanding. The research of image segmentation has been received more attentionin the past forty years and tens of thousands of different algorithms have been raised.Although most of these algorithms in varying degrees, with some success, but theimage segmentation problem is far from settled, the research in this area are stillfaced with many challenges. So far, no one segmentation method for any imageshave been made consistent with good effect, is not all algorithms can get equallygood results in an image. Also, the lack of a unified theoretical system, nor a way toguide us in how to select the appropriate segmentation algorithm according to theimage. Therefore, to solve some practical image segmentation, we are often difficultto find a suitable algorithm. Even find some of the algorithms, each algorithmapplies only to parts of the image.In view of this, the main contents include: the color space for color imagesegmentation and summarized analysis of the characteristics of various types ofcolor space. Summarized and analyzed for the current typical color imagesegmentation method, and then for those content proposed in this articlesegmentation method. This article segmentation algorithm is: the first natural colorimage to select the suitable color space (we select RGB color space), and thenfiltering and eight neighborhood region growing, preliminary subdivision plans inthe region growing algorithm, we choose a color image take the grayvalue(brightest)point as the growth of the initial seed points. Filtering processing weused the median filter, the tradition of most of the filtering method using Gaussian filtering. A disadvantage of the Gaussian filtering is it can’t eliminate noise sampleand will lead to regional segmentation image edge fuzzy inaccurate. And medianfilter is nonlinear filtering, it can also reserves of image edge character based on theindependent eliminate noise realization. And after a median filter treatment of eachpixel actually represents the pixels on the neighborhood mean, this cautious clusterprovides great convenience. The last of the clustering we adopt K-means clustering,the growth of the region after graph clustering analysis to obtain the integrity of thewhole image segmentation. The experimental results show that the method has goodsegmentation performance.