Pixelwise Estimation of Signal-Dependent Image Noise Using Deep Residual Learning

Figure 7

Noise estimation results on a real linear RGB image. In general, the bright part in images suffers from stronger noise, which is in accordance with the noise model (1). The input images are 4K and need to be sliced into patches to process. Therefore, there are some boundary effects in the estimation results. Estimation results are linearly scaled for visualization. (a, b) Real linear RGB image. (c, d) Estimation.

(a)

(b)

(c)

(d)