Study Of Micro-cell Image Mosaic Fusion And Characteristic Cell Extraction

The recognition of the micro-cell image is very important in image understanding and processing. Under the premise of resolution,only local part of the cell image can be obtained in each operation for the depth and vision-limit of microscope. So the focal distance and slide need to be adjusted frequently for better observation and analysis. It will not only bring over-sized work and low-efficiency to the researchers but also a low data-precision and credibility with great risk to the scientific research and productive work. With the development of the elector-optical detection, machine vision and digital image processing, the technology of micro-cell image processing has become a new direction: the fusion-technology is used to obtain a image with all structures clear in different focusing planes; the splicing-technology is used to get a big field-vision image;some related segmentation-technologies are used to extract specified objects for better information analysis.Image splicing technology and Harris vertex extraction algorithm have been studied based on fusion technology,subsequently, a modified method for Harris algorithm and robust vertex match was proposed. Finally, some progresses achieved with the modified FCM cluster algorithm used in the extraction of colored-cell image for specified information were as follows:(1) Image fusion. The low frequency coefficient fusion rule: region-energy combined with weighted-average operator and high frequency coefficient fusion rule: partial gradient were respectively proposed have been proposed based on the wavelet transformation theory with some evaluating-indicators.(2) The improvement of Harris vertex extraction algorithm. The gaussian function was replaced by central B spline function to improve the anti-chirp nature and approaching-signal ability of Harris algorithm; Meanwhile, Auto-adapted method was used to control the vertex number and enhance the distributed uniformity for better vertex match.(3) Vertex match. The robust vertex match was proposed based on the spatial relative and supporting relation of vertex to modify the result obtained by the normalized mutual correlation method for reducing match mistakes, increasing match precision and then establishing the inversion matrix. (4) Image segmentation. The modified FCM cluster algorithm was used to divide the colored image and extract specific objects for information analysis.