武汉理工大学数字图像处理英文课件 Chapter2（9）.pptVIP

Image Histogram Statistical Moments Normal (Gaussian) Distribution Definitions of Variance Matlab Statistics Histogram Moments Image Contrast Image Modulation Matlab Function imstats Example * The histogram of an image plots the frequency of occurrence (count) of gray levels for each gray level present in the image. For a continuous image, the gray scale is partitioned into equally-spaced bins and each pixel is assigned an appropriate bin index (gray level). Each bin counts the number of pixels in the image having the corresponding bin index. In matlab, the function imhist can be used to generate and display image histograms. ? im = imread(kitchen2.tif); ? imshow(im); ? imhist(im); One interesting observation in this example is that the gray levels are not uniformly populated. Mean Variance Skewness Kurtosis Skewness measures the asymmetry of a distribution. It is positive if there is a long tail to the right and negative for a tail on the left. Kurtosis measures the sharpness of the distribution relative to a Gaussian distribution. Negative kurtosis implies that the peak is broader than a Gaussian. Positive kurtosis means that the distribution is sharper than a Gaussian. Skewness and kurtosis are sensitive to outliers, pixels with gray levels far removed from the majority distribution. standard form normalized (Gaskill) x = linspace(-2,2,201); y = exp(-pi*x.^2); plot(x,y,k); sigma = 1/sqrt(2*pi) xr = [sigma sigma]; yr = [0 exp(-0.5)]; hold; plot(xr,yr); plot(-xr,yr); hold off; g(x) with one-s points marked There are two common textbook definitions for the variance s2 of a data vector Z: where m is the mean and N is the number of elements in the sample. The two forms of the equation differ only in N versus N-1 in the divisor. If Z is a random sample of data from a normal distribution, the expression with N-1 is the best unbiased estimate of its variance. The expression involving N is often used in the definition of the moments of a proba