| Summary: | 碩士 === 臺灣大學 === 電信工程學研究所 === 98 === Images and videos are usually corrupted by noise when they are transmitted through communication networks and received from the sensors of antenna. Since the interference reduces quality of the multimedia, we consider the algorithm for noise reduction and design it to decrease the influence of noise. However, the researches for image denoising method are still a hard challenge because it is difficult to distinguish between edge texture and noise component since both of them belong to high frequency component. Therefore, many proposed algorithms which have an outstanding performance in the noise reduction but it is easy fail in avoiding artifact.
Bilateral filter was proposed by C. Tomasi and R. Manduchi in 1998. The filter combines domain filter and range filter which measure geometric closeness and photometric similarity between pixels respectively to remove out noise for edge preserving in the spatial domain. Conventionally, bilateral filter is applied to denoise for grayscale image. This thesis extends the concept to color image for noise reduction. Theoretically, it has a better performance because it considered color distortion. The experiment shows the performance which successfully indicates our algorithm in noise reduction and preserving edge simultaneously.
Bilateral filter is applied to many applications for image processing. In addition to image noise reduction, it is also applied to contrast management of high dynamic image and video noise reduction. We implement the filter based on 2D mask and 3D mask to video noise reduction and compare their outcome. Typically, the volumetric bilateral has better performance and doesn’t produce artifact.
|
|---|
