A method is proposed, which uses the digital image processing technique to identify cloud boundaries from scanning Ka-band (~35.29 GHz) radar imagery dataset. In this method, a cloud is considered as an uninterrupted region of radar echoes with radar reflectivity higher than -34 dBZ and area greater than 3 km ² . The proposed algorithm involves 1) conversion of radar RGB image to grayscale by removing white background and noise, 2) identification of cloud boundaries by canny edge detection, and 3) estimation of cloud cross-section area (CCSA) and cloud top height (CTH) based on the pixel width. This method is effectively applied to Ka-band radar images collected over Mandhardev, a high altitude scanning station in the Western Ghats (WGs), India to derive CTH and CCSA. CTH distribution shows three peaks, one at about 2 km with others at about 7-8 km, and 12 km. The cloud occurrence shows an apparent diurnal variation with a maximum in the afternoon hours while a semidiurnal variation is observed in CCSA. The proposed method shows consistent statistics with Global Precipitation Measurement Ka-band radar observations and thus suitable to build robust cloud climatology over the WGs. Such cloud statistics are essential to validate the representation of clouds in weather and climate models.