Denoising camera data - Emerging Trends in Image Processing, Computer Vision, and Pattern Recognition - page 12

(8)
(9)
We use a threshold δ that depends linearly on the signal variance, which is, as described in
Section 2 , known from measurements.
(10)
Table 7 shows the results of the SA-DCT denoising with additional temporal denoising and
Table 6 shows the result of the single-image SA-DCT without temporal denoising. The tem-
poral denoising step leads to a higher PSNR and effectively reduces the flickering of the video
sequence. The visual quality of the video sequence is therefore clearly higher with the tempor-
al denoising step.