数字图像处理课件(冈萨雷斯第三版)英文翻译课件.ppt

Dynamic range compression The gray range of the original image is too large, so it needs to be compressed to the extent of the display. For example, t = C, log (1+|s|). C is a scaling constant. s t Ash segmentation Bring out the gray of a part. s t histogram transformation Histogram and accumulation histogram The grayscale statistics of images are called histograms. Suppose the image has N pixels, and the gray scale is k = 0,1,... L-1. SK is the gray level of the K level pixels. If the pixel has NK in the image, then the frequency of the pixel is To standardize, limit SK to 0 to 1. When k runs all the way to 0,1,... When L-1, a function graph is given, called histogram. The histogram is a discrete function of 1D. The accumulation histogram is the accumulated histogram generated by the sum of the previous K grades 。 L-1 L-1 Ps(sk) EH(sk) k k Histogram equalization Let PS (SK) = nk/n, 0 SK 1, k=0,1,... L-1., that is, grayscale SK pixels, is PS (SK) in all pixels . Set EH(sk) is a monotonically increasing function, which is equal to the proportion of pixels below sk in gray scale. You can write inverse functions: sk tk sk tk Histogram calculation Histogram equalization of the calculation: assume that the original image has L gray, you may remember sk is grayscale k. Calculate the frequency in turn p(k)＝nk/n, k=0,1,…,L-1. Cumulative histogram EH(k), k=0,1,…,L-1. The approximate values of tk are calculated by the lower form ： Because the TK between 0 and 1, tk should be mapped to the [0, an integer in the range of L-1], so L-1 multiplied by tk after rounding, namely [(L-1) tk]. From this, the gray scale mapping relation of gray k to [(L-1) tk] is determined 的pk Histogram specification Specifies that the histogram of an image becomes a specified histogram. The idea is to turn them into equalized histograms first, and then invert the first step to use them . The transformation inversion needs to determine the corresponding rules. SML selects the lower