How to compare normal distribution value of one RGB image with another?

Question

Jaanu 2011-9-11

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/15578-how-to-compare-normal-distribution-value-of-one-rgb-image-with-another

I am very new person to matlab. For my Msc project, I calculated normal distribution value of an RGB image.Now i need to compare this normal distribution value with another normal distribution value of a RGB image.

I use the following coding to calculate Normal distribution.

x=imread('face7.jpg');

y=x(:);

z=double(y);

mu=mean(z);

sigma=std2(z);

p=normpdf(z, mu, sigma);

w=norminv(p,mu,sigma);

Plz help me it's urgent.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Image Analyst 2011-9-14

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/15578-how-to-compare-normal-distribution-value-of-one-rgb-image-with-another#answer_21523

Who says that the three histograms (red, green, and blue) of a face image are in any way, shape, or form Gaussian? And if it is, what is the chance that a difference face image is? Maybe you should try a different/better image comparison method, something like subtraction or PSNR for super simple but not-too-good ways, or more sophisticated methods like in sections 12, 13, and 14 here: http://iris.usc.edu/Vision-Notes/bibliography/contents.html.

Or look at the color frequency image: http://www.mathworks.com/matlabcentral/fileexchange/28164-color-frequency-image

Or just create a feature vector made up of a whole bunch of measurements like mean colors, spread in colors, number of edges, etc. and compare feature vectors.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

Answer 2

Wayne King 2011-9-11

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/15578-how-to-compare-normal-distribution-value-of-one-rgb-image-with-another#answer_21171

Hi, I'm not sure what your ultimate goal is, but you can use normfit() from the Statistics Toolbox to obtain estimates for mu, sigma, and confidence intervals for these parameters. I'm assuming you have already determined that these data are well-approximated by a normal distribution. You may want to assess that first with something like normplot().

Of course, if you are just interested in a hypothesis test about the two distributions, there is kstest2 and ttest2.

Wayne