영상에 포함된 피쳐의 방향성을 적용한 잡음의 분산 추정과 시그마필터

초록

Although video devices have evolved along with the development of IT technology, it is inevitable that noise will occur in images. Noise blurs image quality, thus accurate information is not available. Since it is practically impossible to obtain images without noise, it is important to effectively eliminate noise before using the images. In this study, after adding Gaussian noise to the images, we consider orientation using the block approach to detect image features and noise. The problem of detect image features and noise can be expressed as homogeneity of variance test, and feature statistic is defined by using the parameter method, Bartlett test. Using STD estimations for local block, STD estimations by feature orientation and feature statistics, we propose estimation algorithm of noise variance. In addition, noise is eliminated by applying the estimates obtained by the proposed algorithm to the sigma filter. As the simulation results, the efficiency of noise reduction of proposed method is excellent regardless of the level of noise. Also, It can also be visually verified.

키워드