Research On Hough Transform For Microscopic Image Segmentation

Hough transform becomes the focus of image segmentation for it has the insensitivecharacteristics of noise and local disconnection. However, it still has the question ofcomplex algorithm and lower efficiency. In this paper, do the depth research on Houghtransform for microscopic image segmentation based on describing the principle of Houghtransform. The main contents are organized as follows:Chord midpoint Hough transform method is employed to detect cell targets byanalyzing the probability of ineffective sampling. The composition of each point group thatused to determine the circle parameters reduces from three points to two points, so that itcan reduce the probability of ineffective sampling. Furthermore, coordinate relationshiplimited is proposed to reduce calculation, and ineffective accumulation can be well solved.Because of the complexity of the cells themselves and lighting factors, it leads thefictitious information in edge detection image. Moreover, most Hough algorithm generallyintroduce edge detection operator of the spatial domain to pre-process, and these operatorsare very sensitive to noise. For these conditions, a morphological holes filling approach isused to fill the binarization image in order to reduce the fictitious edges appearance.Secondly, an improved morphology edge detection method is employed to extract edges,and so it can reduce the noise effectively and obtain detailed edge accurately. It caneffectively solve the question that non-circle points are extracted as the edge points.A color image can provide more information than a gray image. The HSI space whichbetter meets human visual characteristics is introduced in this paper. Process S and I imagerespectively to enhance the brightness of the image. And then, it is as far as possible todisplay cells and background as different colors by using pseudo-color processing. By HSIspace and pseudo-color processing, it is easy to get more and accurate edge information.There is not any filter to lead edge information loss. On these bases, Hough transform isintroduced, and cells can be detected accurately and rapidly.Three algorithms in this paper are tested by Matlab simulation experiments. Resultsindicated these algorithms can extract circular targets more accurately and faster, andactual microscopic cell image segmentation experiments also prove their effectiveness.