首页 > 代码库 > 不同分割方法应用的彩色空间
不同分割方法应用的彩色空间
人类的眼睛能识别上千种颜色和强度但是只能识别two-dozen（24？）种灰度，经常利用灰度提取不到目标但是用颜色却能提取，所以颜色在模式识别和计算机视觉中有着很重要的作用。但是应用在彩色图像上的分割方法不如在灰度图像上的分割方法多。现在来说应用在彩色图像上的分割方法一般如下所示：
Fig. 4.
Full-size table
Commonly used color image segmentation approaches.
彩色图像可以看作一种特殊的多光谱图像，任何可以用在分割多光谱图像上的分割方法都可以用来分割彩色图像。在分割复杂纹理的图像时比较困难，纹理仍被看作图像分割算法中的重要的问题，关于纹理的变化上的讨论比关于颜色的表示多，以及如何从纹理的变化中提取目标的特征都是要解决的问题。分割方法被分为四大类：基于像素的分割，基于区域的分割，基于边缘的分割，基于物理学的分割。
基本上大部分的应用于灰度图像上的分割经过一定变化都可以应用于彩色图像的分割。但是其中一个问题就是如何把颜色的信息来代表一个像素的整体信息。当颜色被三个分量来表示时，颜色的信息就变得过于分散而导致一个彩色图像单纯的变为多光谱图像丢失了许多人类能感知到的颜色信息。因此如何选取颜色的表示方式对图像的分割有着很重要的意义。每种颜色的表示方式都有着自己的优点和缺点。至今没有一种颜色空间能应用到分割每种图像中去。
现有的大部分的分割算法里，对分割得到的区域定义大多是基于颜色的相似性，因此造成分割如有高光，阴影，以及具有复杂纹理而产生不均匀颜色的物体表面的图像时比较困难。用HSI彩色空间可以一定程度上解决这种问题除了在低饱和度的情况下导致的色调过低的情况。
Table 1.
Monochrome image segmentation techniques
| Segmentation | Method description | Advantages | Disadvantages |
|---|---|---|---|
| technique | |||
| Histogram thresholding (mode method) | Requires that the histogram of an image has a number of peaks, each corresponds to a region | It does not need a prior information of the image. | (1) Does not work well for an image without any obvious peaks or with broad and flat valleys |
| For a wide class of images satisfying the requirement, this method works very well with low computation complexity | (2) Does not consider the spatial details, so cannot guarantee that the segmented regions are contiguous | ||
| Feature space clustering | Assumes that each region in the image forms a separate cluster in the feature space. Can be generally broken into two steps: (1) categorize the points in the feature space into clusters; (2) map the clusters back to the spatial domain to form separate regions | Straightforward for classification and easy for implementation | (1) How to determine the number of clusters (known as cluster validity) |
| (2) Features are often image dependent and how to select features so as to obtain satisfactory segmentation results remains unclear | |||
| (3) Does not utilize spatial information | |||
| Region-based approaches | Group pixels into homogeneous regions. Including region growing, region splitting, region merging or their combination | Work best when the region homogeneity criterion is easy to define. They are also more noise immune than edge detection approach | (1) Are by nature sequential and quite expensive both in computational time and memory |
| (2) Region growing has inherent dependence on the selection of seed region and the order in which pixels and regions are examined | |||
| (3) The resulting segments by region splitting appear too square due to the splitting scheme | |||
| Edge detection approaches | Based on the detection of discontinuity, normally tries to locate points with more or less abrupt changes in gray level. Usually classified into two categories: sequential and parallel | Edge detecting technique is the way in which human perceives objects and works well for images having good contrast between regions | (1) Does not work well with images in which the edges are ill-defined or there are too many edges |
| (2) It is not a trivial job to produce a closed curve or boundary | |||
| (3) Less immune to noise than other techniques, e.g., thresholding and clustering | |||
| Fuzzy approaches | Apply fuzzy operators, properties, mathematics, and inference rules (IF– THEN rules), provide a way to handle the uncertainty inherent in a variety of problems due to ambiguity rather than randomness | Fuzzy membership function can be used to represent the degree of some properties or linguistic phrase, and fuzzy IF–THEN rules can be used to perform approximate inference | (1) The determination of fuzzy membership is not a trivial job |
| (2) The computation involved in fuzzy approaches could be intensive | |||
| Neural network approaches | Using neural networks to perform classification or clustering | No need to write complicated programs. Can fully utilize the parallel nature of neural networks | (1) Training time is long |
| (2) Initialization may affect the results | |||
| (3) Overtraining should be avoided |
Full-size table
Table options
Table 2.
Characteristics of color spaces
| Color space | Advantages | Disadvantages |
|---|---|---|
| RGB | Convenient for display | Not good for color image processing due to the high correlation |
| Can be used to efficiently encode color information in the TV signal of American system; | Correlation still exists due to the linear transformation, though not as high as RGB | |
| YIQ | Partly gets rid of the correlation of RGB; | |
| Involves less computation time; | ||
| Y is good for edge detection; | ||
| Can be used to efficiently encode color information in the TV signal of European system; | Correlation still exists due to the linear transformation, though not as high as RGB | |
| YUV | Partly gets rid of the correlation of RGB; | |
| Involves less computation time | ||
| Partly gets rid of the correlation of RGB; | Correlation still exists due to the linear transformation, though not as high as RGB | |
| I1I2I3 | Involves less computation time; | |
| Can be useful for color image processing | ||
| Based on human color perception; Useful in some cases where the illumination level varies, because hue is invariant to certain types of highlights, shading, and shadows; | Nonremovable singularity and numerically unstable at low saturation due to nonlinear transformation | |
| HSI | Hue can be useful for separating | |
| objects with different colors | ||
| The individual color components are independent on the brightness of the image; | Very noisy at low intensities due to nonlinear transformation. | |
| Nrgb | Convenient to represent the color plane; | |
| (Normalized rgb) | Robust to the change of the illumination | |
| Can control color and intensity information independently; | Have the same singularity problem as other nonlinear transformations do | |
| CIE spaces | Direct color comparison can be | |
| (L?u?v? or L?a?b?) | performed based on geometric | |
| separation within CIE space, and | ||
| efficient in measuring small color | ||
| difference |
Table options
总结：
至今没有一种图像分割的理论是通用的，也没有一种颜色空间是通用的，大部分灰度图像的分割方法都能够扩展到彩色图像的分割中去，例如阈值分割，聚类，区域生长，边缘检测，基于模糊的方法等，它们可以应用到彩色图像的一个颜色分量当中去，然后通过一些方法将每个分量的结果结合在一起，得到最终的分割结果，然而有一个问题，如何使用这些颜色的信息作为每一个像素的整体，当彩色图像变为多光谱图像时人类感知的颜色信息会有丢失。另外一个问题就是如何选择彩色空间来表示一幅图像，因为每种颜色空间都有自己的优点和缺点。
不同分割方法应用的彩色空间
声明：以上内容来自用户投稿及互联网公开渠道收集整理发布，本网站不拥有所有权，未作人工编辑处理，也不承担相关法律责任，若内容有误或涉及侵权可进行投诉： 投诉/举报 工作人员会在5个工作日内联系你，一经查实，本站将立刻删除涉嫌侵权内容。