The Human Visual System Model-based Image Enhancement Algorithms And Partial Differential Denoising

This thesis emphasizes on two pivotal parts of digital image processing: image enhancement and image denoising. Image enhancement technique was first born with the requirements of spatial imaging and medical imaging, and then became the basic issue in digital image processing. Nowadays, with the prevalence of Digital Cameras, Digital Videocon and other digital devices, a tidal wave of image enhancement techniques comes forth for a second time. In this thesis, first, color spaces and color models are introduced, the Human Visual System (HVS) is explicated and lateral inhibition mechanism is emphasized. Then various image enhancement algorithms are analyzed, and two automatic image enhancement algorithms are advanced: One is image enhancement with adaptive non-symmetric Retinex-based filter;the other is a highly robust real-time image enhancement algorithm which combines global compression and local processing according to the features of the input images. The results show that both of the two methods effectively enhance the images' qualities and are robust enough to various images, including complex scenes with highly varying luminance.Image denoising is another significant part of digital image processing. For the limitation of sensor devices and infection of communication channels, received images are frequendy disturbed by noise. Therefore, noise elimination with maintenance of image features is a practical research theme. In this thesis, three typical methods for image denoising are introduced, which are mathematical morphology, Gaussian smooth filter and TV (Total Variation) model. The results show that image denoising with TV digital filter performs well and maintenance the edge information of the images.