英文附录

  • 格式:docx
  • 大小:46.54 KB
  • 文档页数:28

下载文档原格式

  / 28
  1. 1、下载文档前请自行甄别文档内容的完整性,平台不提供额外的编辑、内容补充、找答案等附加服务。
  2. 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
  3. 3、如文档侵犯您的权益,请联系客服反馈,我们会尽快为您处理(人工客服工作时间:9:00-18:30)。

Digital Image Segmentation Technology and Its Development

Abstract: Digital image segmentation is a digital image processing difficulties and focus on technology, starting from the basic concepts of digital image segmentation, summarized the main method of digital image segmentation, review the image segmentation quality evaluation, and finally discusses the digital image segmentation problems and trends.

Keywords: digital image; segmentation; quality evaluation;

Introduction

Digital image segmentation is a digital image processing is a key technology, it is usually for further image analysis, recognition, compression and other aspects of image pre-processing, segmentation accuracy of its direct impact on the effectiveness of subsequent tasks, so has very important significance. Since the 1970s has been highly valued by people, has been presented thousands of segmentation algorithm, but there is no general theory of division, resulting in a large current segmentation algorithm is proposed for the specific problem, and no one is suitable for all Generic image segmentation algorithm. In addition, there is no segmentation algorithm applied to develop a standard, which brings a lot of practical problems to the application of image segmentation techniques. Has emerged in recent years many new ideas, new methods and improved algorithms. Firstly, the research status of digital image segmentation, segmentation and some classical emerging segmentation method outlined, followed explores the digital image segmentation quality evaluation system standard, the final analysis of the digital image segmentation technology problems and trends.

The basic concept of a digital image segmentation

Digital Image Segmentation refers to a digital image into several mutually non-overlapping, meaningful, area having the same nature. Good image segmentation should have the following characteristics:

(1) out of the regional division for certain properties (such as gray, texture) in terms of similarities, within the region is connected and not too many holes.

Significantly different (2) adjacent to the divided area is based in nature.

(3) zone boundary is clear.

Most image segmentation method is only partially meet the above characteristics. If stressed divided region, with the nature of the constraint is partitioned area is prone to a lot of holes and irregular edges; if to emphasize the nature of the differences between different regions

was significant, could easily lead to merge different regions. The specific treatment of different image segmentation methods always find a reasonable balance between the various constraints.

The main method of digital image segmentation

From the beginning of the fifties and sixties of last century, scholars have been keen to image segmentation techniques. So far, thousands have been proposed for image segmentation algorithm, according to these algorithms for image processing features, it can be divided into the following categories methods.

Thresholding Method

Thresholding method as a common area of parallel technology, which by setting the threshold, the pixel grayscale divided into several categories, in order to achieve image segmentation. Because it is the direct use of gray-scale characteristics of the image, and therefore easy to calculate concise, practical. Obviously, the key threshold segmentation method and difficulty is how to get an appropriate threshold, and the practical application of the threshold set susceptible to the impact of noise and light brightness.

In recent years, on the threshold segmentation method are: the principle of maximum correlation method selection threshold, image-based topology steady state law, GLCM method, entropy method, peaks and valleys analysis method.

In which the adaptive threshold method, the maximum entropy method, fuzzy threshold, the threshold between-class method is the more successful of several algorithms to improve the traditional threshold method. More often, the choice of threshold will be integrated use of two or more kinds of methods, and this is a trend in the development of image segmentation. For example, the image histogram seen as a combination of Gaussian distribution with adaptive directional selection method of Gaussian orthogonal projection decomposition method, a better fit of the multimodal characteristics of the histogram, resulting in a more accurate segmentation effect.

The main defect threshold method is that it takes into account only the gray scale information, while ignoring the spatial information of the image. Simple image processing for one or the other (eg some binary image processing) is valid, but the image does not exist significant difference in gray scale or gray scale value range of the object overlaps a larger image segmentation difficult accurate segmentation.

Edge-based image segmentation method