The Study On The Image Thresholding Segmentation Algorithm

Image segmentation is very important in the field of digital image processing, Theprinciple of Image thresholding segmentation is simple, low computation, segmentationresults good, so it is common algorithm of segmentation, On the basis of reviewingnumerous existing methods, this article do some research from the aspect of imagethresholding, Main content of the thesis are as follows:First the background and significance of the subject are introduced, summarizes theimage threshold segmentation research situation at home and abroad, then introduced onedimensional histogram, two dimensional histogram, the concept of three-dimensionalhistogram, and several classic image thresholding algorithm.Cumulative residual entropy is a new way of information measurement, it isintroduced into the infrared image segmentation, analysis the principle of maximumcumulative residual entropy thresholding method in theory, that is finding the gray valuewhich make the cumulative residual entropy maximum as the threshold,the maximumcumulative residual entropy threshold method can be understood as transforming thehistogram first,and then solve the threshold in the processed histogram.one dimension histogram is the most simple image model of statistics, the gradientof gray histogram thresholding selection method and the maximum cumulative residualentropy thresholding method is proposed by this article on the basis of one-dimensionalhistogram.The Gradient Gray histogram thresholding selection method make full use ofthe histogram of gradient information, using gradient of the gray histogram process thehistogram,Get the gradient direction of the histogram, and make the gray level whichthe gradient in the gradient direction changed from positive to negative turning point as apeak, from negative to positive turning point as bottom, the turning point which the sum ofthe Negative gradient turn positive gradient is the largest is the threshold.Multidimensional gray histogram consider more information of image than adimension histogram, can get better segmentation results, but while increased thecomplexity of the algorithm, the two-dimensional maximum scatter differencethreshold segmentation method is presented, which not only taking into account thepixel gray value, but also considering the neighborhood pixel gray value, a fast recursivealgorithm is also presented by we, this method can get good balance at the between-classscatter difference and within-class scatter difference by adjusting the parameter in thesegmentation. In the light of the shortcomings of the multidimensional histogram thresholding have large computation. The relation of the thresholding between the onehistogram and the multidimensional has been analysised, Fast algorithms is presented,which is on the basis of the one dimensional histogram, take the three dimensional Otsu’sthresholding for example, first calculate the threshold value of1-D Otsu,then calculatethe threshold value of3-D Otsu in the range of1-D Otsu threshold.The unified multi-dimensional histogram form is Proposed, a weightedhistogram-based threshold segmentation method is proposed on the basis of the unifiedform of multi-dimensional histogram, which segmentation results can made a goodcompromise in preserving image details and Denoising by adjusting the weightingfactorFinally, summarizes the image threshold segmentation algorithm and prospects,improved threshold segmentation algorithm of histogram and put forward, there are threemain points, a establishing a reasonable histogram model, the second is the proposed newcriterion of thresholding segmentation algorithm, third, improve the calculation speed ofdimensional thresholding...