Study On Image Segmentation

Segmentation is one of the most difficult and important steps in digital image processing.Segmentation accuracy determines the eventual success or failure of computerized analysis procedures.For this reason,image secmentation has been widely investigated for more than 40 years,and hundreds of algorithms have been presented in the literature.Although those algorithms are to some extent successful,image segmentation is still far from been solved.There is no single method which can be considered good for all images,nor are all methods equally good for a particular type of image.Moreover ,due to the lack of systematic theory,There is no idea to guides us how to choose appropriate algorithms for different images.There have huge number of existing image segmentation algorithms,also have different classifications.In particular,we will be divided the algorithm into two classes:1.based on image segmentation data processing,data-driven strategy;2. Based on the model in image segmentation .This article focuses on the first type. The first type can be divided into the following categories:1. Based on image threshold segmentationImage threshold segmentation is a widely used image segmentation techniques. The image can be viewed as the combination of two kinds of parts with different gray level on the basis of difference between image object and background in gray level. By selecting a suitable threshold, the every point of the image can be determined to belong to either object or background, thus getting a image with binary value. This kind of method has some merits such as easy data compressing, small storage size, and simplifying the subsequent analysis and treatment.2. Based on district-trackingdistrict-tracking is seeking the pels group with similarities, which is corresponding to some practical plane or object. According to the rules of uniform characteristics (this uniform can be gray level, color ness, texture, grads and others), the pels with similar characteristic will be gradually increased, this is called district increase. Some uniformity estimating function are used to estimate the uniformity of those pels. If the result is real, the district is not going on increase until the result becomes false. The common used methods include district segmentation, district growing method and district segmentation -merging method.3. Based on edge detectionEdge in the future and prospects, between foreground and background, the gray region and between regions of rapid change, not part of a row, according to their characteristics to segment the image is an important segmentation algorithm, a classic segmentation.Image segmentation technique in many high-end image processing and computer vision field of application software programs take a very important step, the importance of image segmentation to enable a quantitative evaluation of image segmentation is imperative. Reliable evaluation method has two characteristics: 1, can effectively compare different segmentation methods, for any segmentation method to different parameters, the final segmentation can get the same conclusion. 2, can choose a good segmentation, segmentation results are good segmentation helps different follow-up data-processing steps to obtain high-level image processing results.About evaluation of image segmentation:Generally believed that There are two ways to evaluate algorithm of the segmentation by analysis and experiment . Therefore, the segmentation evaluation methods can be divided into analysis and experimental two categories. Analysis is Direct analysis of segmentation own principles and functions of the experimental method is by Segmentation results to evaluate the algorithm. Divided by the actual results of the analysis to assess the segmentation is meaningful. There have many experimental evaluation of the segmentation methods.