A multiresolution segmentation method for tree crown image using wavelets

Tree crown image segmentation is a major step of many applications, such as crown transparency estimation, virtual plant, and so on. Varying image characteristics caused by differing levels of foliation and fluctuating lighting conditions present a unique challenge for tree segmentation. For example, small leaves surrounded by bright sky, small openings among dense canopy, and mixed pixels. This paper proposes a new segmentation method, based on wavelets and Markov random field model, and considering color information (green band). The morphological wavelet transform provides a powerful transform domain for modeling singularity-rich structure images. The Markov random field model provides a good tool for distinguishing between different textures. The new method provides both lower classification error rates and better visual results, especially where crown objects may have complex backgrounds.