Tresholding of Color Images Based on Means

Thresholding is one of the most stable methods for image clustering. Nevertheless, determination of proper thresholds for every image is difficult process in terms of computation cost. Furthermore, the methods developed so far have only been applied for gray scale images. In this study, an algorithm which works with both gray scale and color images has been developed. Histogram data of each channel and recursive means have been employed for threshold estimation. Initially, global mean was calculated. Subsequently, means determined from previous steps was reused. Thus, threshold number was increased with each step. Accordingly, the RGB color space was partitioned with the thresholds information obtained for each channel. The pixels located in each sub cube were assigned into the same cluster