Noise reduction using variance characteristics in noisy image sequence

Abstract

In this paper, a new of noise reduction algorithm using variance characteristics of noise is presented. Estimated noise power and sum of absolute difference (SAD) employed in motion estimation are used to determine temporal filter coefficients. A noise measurement scheme using the correlation between the noisy input and the noise-free image is applied to accurate estimation of noise power. Experimental results support that the proposed noise reduction method efficiently removes noise.