2.1. Flower segmentation
The first step in flower classification is to segment the flower image by removing the unwanted background region.
In general, autonomous segmentation is one of the most difficult tasks in image processing. Flowers in images are often
surrounded by greenery in the background. In order to avoid matching the green background region, rather than the desired
foreground region, the image has to be segmented. To segment the flower image, we use a semi-automated threshold-based
segmentation algorithm [11]. A given image is transformed to the HSV plane and an intensity histogram corresponding
to each channel is extracted. The histogram intensity values corresponding to two dominant regions belonging to the
background and the flower are identified. Based on these intensity values, the flower is segmented. Fig. 2 shows the results
of flower segmentation using the threshold-based method on a few sets of images with a cluttered background.