A fluctuating solar irradiance field instantly influences the output power of PV installations. The major impact factor for small-scale temporal and spatial solar irradiance variability of insolation are clouds. By means of automatic sky imagers an operational monitoring and short-term forecasting of the cloudy atmosphere has been performed. A processing chain for analyses and forecasts of cloud shadow distributions from full sky images has been developed. The original picture (Fig. 1a) is processed by a calculation of the ratio of red and blue color (RBR) and the greyscale values. We found the fixed RBR criteria for the detection of cloudy pixels in the image is a simple approach with a weak performance for clear sky cases. Due to the assessment of the sun disc and its surrounding saturated area as cloudy pixels, errors in the total cloud cover of up to 40 % occur. A modification to the RBR (Fig. 1b) based on clear sky characteristics and surface irradiance measurements has been done to improve the retrieval performance around the bright sun region. An empirical determined threshold is applied to this red-blue-ratio RBR to classify each pixel as either cloud or sky (Fig. 1c). Using the lens mapping function of the sky imager leads to the undistorted distribution of the clouds. The final projection of the clouds shadows is calculated by using measured or estimated cloud base height, the position of the sun and geometry (Fig. 1d). For a typical cloud base height of 1000 m we found a pixel resolution of the shadow at the ground between 3 m (centre of image) and 15 m (horizon).