GENERAL CONCLUSIONS



  • Developed an algorithm for estimating the scenic beauty rating of forestry images

  • Developed feature extraction modules for various features

    - Edge Detection
    - Fractal Dimension
    - Entropy
    - Sharpness
    - Standard Deviation
    - Compression Ratio

  • Developed an extensive database in conjunction with USDA

  • The best error rate is 36.8% with the classification approach for the RGB + Long-Lines + Entropy model

  • The best correlation is 0.65 using all features

  • Software, data and documentation are available from the ISIP web site at: http://isip.msstate.edu/resources/technology/research/isip/projects/1997/sbe_imaging/