Adaptive image restoration based on hierarchical neural networks

We propose a new approach to adaptive image regularization based on a neura l network, hierarchical cluster model (HCM). The HCM bears a close resembla nce to image formation. Its sparse synaptic connections are effective in re ducing the computational cost of restoration. We attempt to achieve adaptiv e restoration by assigning entries of a novel, optimized regularization vec tor to each homogeneous image region. The degraded image is first segmented and partitioned into smooth, texture, and edge clusters, it is then mapped onto a three-level HCM structure. An evolutionary scheme is proposed to op timize the regularization Vector by minimizing the HCM energy function. The scheme progressively selects the well-evolved individuals that consist of partially restored images, their corresponding cluster structures, segmenta tion maps, and the optimized regularization vector. Experimental results sh ow that the new approach is superior in suppressing noise and ringing at th e smooth background while effectively preserving the fine details at the te xture and edge regions. An important feature of the method is that the empi rical relationship between the optimized regularization vector and the loca l perception measure can be reused in the restoration of other degraded ima ges. This generalization removes the overhead of evolutionary optimization, thus rendering a very fast restoration. (C) 2000 Society of Photo-Optical instrumentation Engineers. [S0091-3286(00)03507-8].