Wavelet Transform And Morphology Image Segmentation Algorism For Blood Cell Image Analysis

Image segmentation is the key step between image processing and image analysis, which is also the truth in medical image processing. The original image is always transformed into a more abstract and more compact form after the processing of image segmentation, target separation, feature extraction and parameter measurement, which helps the doctor to make a more accurate analysis and diagnosis of the disease. Many kinds of classic segmentation algorithms have already been presented, but no a segmentation method gives the desired segmentation results to all kinds of images. Based on the characteristics of candidate image, it is always a new important way combined with various methods to get a segmentation model in line with the specific image features. The hybrid algorithm can always improve the effect of image segmentation by taking advantage of different methods. Thus the main work done in the thesis is as follow.(1)The multi-resolution characteristics of wavelet transform is studied, based on the space and frequency-domain partial transformation of wavelet transform, the cell image is multi-scale refined by scaling and translation operations. Then, an adhesive cell image is segmented based on mathematical morphology, and the result is not satisfactory with a serious over-segmentation phenomenon.(2)An image segmentation hybrid method is presented based on wavelet transform and morphological. First of all, cells image is decomposed by wavelet transform. The low-frequency components and high frequency components is extracted from the original image at different scales and different directions. Then the image details after wavelet decomposition are processed using mathematical morphology operators such as expansion, corrosion and so on. Due to over segmentation, the region merging is used based on similarity property of the region gray. Finally, the target image is acquired by the inverse wavelet transform with the wavelet coefficients at different scales.(3)Taking adhesive cell images as research object, the hybrid algorithm is implemented by the program of C language in VC 6.0 environment. The experimental results indicate that this algorithm has strong anti-noise ability, high accuracy on edge location, good edge continuity, cell division complete, and effective weak edge detection, and it can segmented the kind of the adhesive cell image successfully.