Özet:
This thesis focuses on three main issues; namely, (i) image acquisition of cancer tissue, (ii) image stitching and (iii) image processing for cancer detection. An algorithm and a related circuity are implemented to create images from signals obtained by a photomultiplier tube (PMT) for the image acquisition part. Data acquisition card and MATLAB are used to acquire signals and create images. The current generated by photomultiplier tube, coming signal is converted to a voltage signal by the designed circuit. Afterwards, implemented algorithm transforms this voltage to a meaningful image and then displays on the screen. The second issue is image stitching. Image stitching is a method which combines a sequence of partially overlapping images into a merged image using an appropriate transformation. Generally, in image stitching, features of images are detected, described and stored to nd matching points. Next, two images are combined with respect to the similarity of matching points in the light of implemented transformation. Image stitching algorithms in the literature are investigated. Final section composes of image processing for cancer detection. Two di erent kinds of novel classi ers are developed to detect cancer. The rst one depends on the analysis of root mean square error (RMSE) owing to line plot of center row of images in the Fourier domain. Unlike healthy tissue, cancerous tissue has an irregular structure because of uncontrollable growth in the cells. Hence, it has been noticed that line plot of the center row has even, symmetric function characteristic if it is healthy tissue image. The other classi er depends on histogram-based threshold technique. It has been realized that cancerous tissue images have darker pixel values than healthy tissue images. Therefore, histogram is considered as a tool to detect cancer.
