SUMMARY
Segmentation of leaf region from background is one of the essential pre-processing steps required in the Plant Leaf Image Processing. This paper proposes an innovative segmentation approach for extracting color leaf region from the healthy or infected plant leaf image with background using an enhanced automatic GrabCut algorithm that does not take any input from the user. In this method, first GrabCut algorithm was applied on the original image. The algorithm removes background but shadows remain in the resultant image which may cause misinterpretations in further processing steps. Hence, the shadows in the image were removed by thresholding ‘a’ and ‘b’ components of CIELAB color space. This step created holes in the infected region, which had similar color as that of shadow, of the leaf image. Hence, the image obtained was binarized and holes were filled with white (foreground) colorizing Flood Fill algorithm. From this binary image containing only leaf region, the color leaf region of the image was filtered. The accuracy achieved was 98%.