Wavelet domain binary partition trees for image segmentation

When performing segmentation using tree based representations it is evident that contiguous objects are often dispersed widely across the tree making controlled selection difficult. This paper demonstrates robust segmentation by the generation of binary partition trees entirely within the wavelet domain. By using spatial frequency in conjunction with colour, a threshold free tree is produced which compels objects to reside inside single branches of the tree by constraining their merging development. Experimental results show the superior performance of the proposed algorithm compared to the existing colour based BPT algorithms.