Research Of Medical Microscopic Image Segmentation Based On OTSU

The obtain of microscopic image has improved medical research into cell level. Shooting pathological tissue slice to get microscopic image makes doctor diagnosis disease more accurately.Computer technology improve rapidly in the medical microscopic image processing applications more and more widely. Many problems in computer and medical image combining has attracted great interest of the researchers, fast and automatic segmentation of medical images is one of them.Immunohistochemistry(IHC) by cell antibody on the color reaction of information extraction is a key technology, biological imaging. In recent years, many in the past can not be diagnosed cancer due to the emergence of IHC technology has been the diagnosis. The maximum between class variance method(OTSU) is a kind of automatic non parametric graph cuts method, it is simple in operation and the segmentation effect is good, widely used in graph cut. There are a lot of negative region based on OTSU IHC image segmentation results, and insufficient to make improvements in noise, the main work:(1) Combined with the improved algorithm of HSV space. There are a lot of negative region of the IHC image is obtained after OTSU segmentation, segmentation is not complete, and rich in noise, which will lead to the statistical errors in the latter part of the medical measurement. The study found that the main existence negative region of the IHC image of H component in HSV space, the main positive area of S component, then puts forward the improved algorithm: first of all,using OTSU as a coarse segmentation of image, and then use the set operations of H components of image in HSV space and S components on the coarse segmentation result is improved.Experiments show that the improved the algorithm is more precise extraction of positive regions in IHC images.(2) The two-dimensional OTSU immunohistochemical image segmentation algorithm. OTSU algorithm is adopted to compute the background and target discrete measure the trace of a matrix to obtain the optimal segmentation threshold, the study found that, when the image of the target and background is similar to the gray, gray histogram may not show peaks and troughs, using the histogram threshold segmentation will lead to wrong. Therefore, in this paper, OTSU is extended to two dimensional, two-dimensional gray histogram of pixel and its neighborhood average gray,was used for extraction of 2D OTSU positive area of rat liver IHC images, and then use the set operations of H components of image in HSV space and S components on the extraction results were improved, achieved better the segmentation effect.(3) Fast implementation algorithm. This paper analyzes the one-dimensional OTSU threshold,the segmented target gray value and average gray scale of background of the average, and on the basis of a fast algorithm for two-dimensional OTSU can better compensate for inadequate;one-dimensional OTSU wrong segmentation, but each of its calculation of the target and background between class discrete measure the trace of a matrix, are required to traversal of the whole image, the huge amount of computation, this paper makes an improvement from the front, a known amount calculated to simplify the calculation of the current position quantity, greatly enhance the efficiency of image segmentation.