In their work, Saitoh et al. [8] describe an automatic method for recognizing a blooming flower based on a photograph
taken with a digital camera in a natural scene. They have also proposed a method for extracting the flower regions. It is
based on ‘‘Intelligent Scissors’’ [9], which find the path between two points that minimizes a cost function dependent on
image gradients. The method works under the assumption that the flower is in focus and in the centre of the photograph
and that the background is out of focus. Under this assumption, the cost between any two points on the flower is smaller
than the cost between a point in the background and a point in the foreground. The midpoint of the image is used as the
starting point to identify the flower region. This method requires no prior color information. Saitoh et al. [10] designed a
flower classification system which extracts features from both flowers and leaves, and used a piecewise linear discrimant
analysis for recognition on a dataset of 34 species each containing 20 sets of wild flowers.