Image Speckle Noise Processing And Edge Characteristics Extracting In Optical Coherence Tomography

Optical coherence tomography (OCT) is a new tomography. Because of its non-destructive, real-time, high resolution, etc, OCT has a bright future in biomedical and materials science. OCT image processing research is carried out in the thesis.The principle and current situation of OCT technology, the hardware and software of OCT system are described. The characteristics of OCT image data, particularly the speckle noise and the effects on OCT image are analyzed, and then OCT image processing methods are investigated.Making OCT image analysis and judgments accurately is a key step in the application of OCT. An algorithm for OCT image feature extraction is proposed, which can effectively help medicals and engineers to process OCT images properly and reach the right diagnosis results.Phase Congruency (PC) algorithm have a particularly function on extracting image features, but it is sensitive to noise, therefore it is only applicable to images with less noises. From the comparison of the speckle noise reduction effects between the double tree complex wavelet transform (DTCWT) and the anisotropic diffusion (AD) algorithms on OCT images, both the two algorithms have some good effects for speckle noise reduction and image quality improvement, and help us better to analyze and judge OCT images.Denoising for improving the image quality has some significance, and further, the edge feature extraction of the image is more valuable to judge and analyze biomedical OCT images. By the combinations of the AD algorithm with, and the DTCWT algorithm with the PC, respectively, from the experiment results of the edge characteristics extraction, although denoising effect of DTCWT algorithm is superior to AD algorithm, but the maintaining of the edge characteristics is not good enough. By combining AD algorithm with modified PC algorithm, a good noise reduction and edge characteristics enhancement, could be achieved, the PC algorithm is more suitable for image pretreatment. These good experiment results may beneficial to establish a database for pattern recognition, and to meet rapid analysis and judgment of OCT images.