Texture Representation and Synthesis Using Correlation of Complex Wavelet Coefficient Magnitudes

摘要: We present a statistical characterization of texture images in the context an overcomplete complex wavelet transform. The is based on empirical observatio ns regularities such images, and parameterized by (1) local auto-correlatio n coefficients each subband; (2) both auto-correlation cross-correlation o f coefficient magnitudes at other orientations spatial scales; (3) first few moments th e image pixel histogram. develop efficient algorithm for synthesizing random im ages subject to these constraints using alternated projections, demonstrate its effectivenes s wide range synthetic natural textures. also show flexibility repres entation, applying variety tasks which can be viewed as constrained synthesis problems. Vision arguably our most important sensory system, judging from ubiquity visual forms communication, large proportion human brain devoted processing. Never