Gene Chip Image Processing Base On Shape Adaptive DCT

Gene chip, the concept of which from computer chip, has been developed for about twenty years. It can get biological information in a large-scale, fast and efficient way. The original information of gene chip is stored in its images. There are no direct signal values of every spot in the images. So image processing and intensity extraction should be done to obtain relevant datas. These are discussed in this thesis.First, in gene chip image preprocessing, different noise models and several general filtering methods are discussed. Then a new method called shape adaptive DCT filtering is introduced. By comparing varying-scale directional kernel estimates, this method adaptively selects, for each point in the image, a set of directional adaptive-scales and its neighborhood. And with the characteristics of DCT, this new method can eliminate noise well. The simulation results prove this.Second, general methods of image segmentation are discussed and simulated. Then the modified K-means cluster algorithm is presented in detail. K-means cluster algorithm does well in large amounts of data. But it has some shortcomings. So K-means cluster algorithm is modified in this thesis. After simulations, it is showed that this method is accurate, stable and fast in gene chip image segmentation.Finally, after extracting and adjusting the spot and background intensity information, data preprocessing is carried out by using the scatter plot and the MA plot. These plots help to find out whether gene expression states increased or decreased.