Basic knowledge in handling the recent review of the image.
Disadvantages: ① to reduce the converted grayscale image, disappearance of certain details; ② some images, there histogram peak, after treatment unnatural excessive contrast enhancement
Gradation unevenness.
1, the principle
Histogram equalization will cause the gray level of false contour merger [] - ostartech - blog Park https://www.cnblogs.com/wxl845235800/p/7429223.html
2, shortcomings
Q: Why histogram equalization histogram is not flat? Why gray levels will be reduced?
A: Histogram gradation level distribution has a uniform probability density, extended dynamic range of pixel values, but reduces the gradation.
1 was found, the histogram equalization is generally not uniform (uneven), but can be extended to the pixel 0-255, it is generally better visual effect of the image;
Explanation:
In class uint8 grayscale image Case , a total of 256 gray level, after equalization, each gray level can be assumed to take, then the original image and the image is completely the same as equalization, if the histogram of the gray-scale image is not flat class uint8 , the image after the histogram equalization is not equalized; on the other hand, if equalization to take certain gray level may not, then the number of pixel level is 0, and thus can be flat of.
Also, if after a certain number of the original image gradation number, the number of pixels corresponding to the equalization stage is a lot, so also it causes uneven.
2 was found, a second histogram equalization processing result with the first result of the histogram equalization same.
Image exactly equal
Why discrete histogram equalization histogram is not flat?
Since the histogram distribution is discrete image, the gray cumulative distribution function is a step function unabated. When the image is mapped to the still take all gray levels, any change does not occur.
If the mapped gray level is less than 256, the histogram will have converted some of the gray level gap . That is, after adjusting the gray level of probability of substantially the same value can not be obtained, so that the generated histogram is not completely flat.
【forward from】
数字图像处理点运算和直方图处理 - 百度文库 https://wenku.baidu.com/view/498367bd5ef7ba0d4a733b67.html
直方图均衡化 - zhulf0804的博客 - CSDN博客 https://blog.csdn.net/zhulf0804/article/details/52770613
为什么离散的直方图均衡化后得不到平坦的直方图_百度知道 https://zhidao.baidu.com/question/338427704.html
直方图均衡化 - Pony_s - 博客园 https://www.cnblogs.com/Ponys/archive/2013/04/30/3051779.html
直方图均衡化 http://www.360doc.com/content/14/1108/11/18306241_423533450.shtml