Denoising algorithms are usually tested on standard test images with artificial white Gaussian noise added. This noise model cannot be applied in the denoising of digital images taken with a single sensor camera because of the signal-dependence of the noise, the demosaicking and the color transformations. We study the noise characteristics with respect to the signal domain. Noise distribution and variance are measured in the raw data and approximated using a Gaussian distribution with a variance linearly dependent on the signal. We evaluate the influence of white balance, debayering and the signal domain and calculate the spatial correlation of the noise. In our experiments we both evaluate the influence of the noise characteristics on human perception and on the performance of denoising methods. Based on a subjective test with 18 participants we can show that the spatially correlated camera noise is more visible than the white Gaussian noise and decreases the visual quality of color image sequences significantly. To evaluate the impact of the noise characteristic on denoising, two state-of-the-art denoising methods are applied to our test data. When the noise is signal-dependent and spatially correlated through debayering the peak signal-to-noise ratio (PSNR) decreases by up to 8 dB. We conclude that it is very important to take into account the correct noise characteristics for increasing the visual quality of color image sequences in future research.