Sparsely Connected Autoassociative Lattice Memories with an Application for the Reconstruction of Color Images

作者：Valle Marcos Eduardo^*; Grande Vicente Daniela Maria

来源：Journal of Mathematical Imaging and Vision, 2012, 44(3): 195-222.

DOI：10.1007/s10851-011-0322-y

• Associative memories
• Mathematical morphology
• Lattice theory
• Artificial neural networks
• Color images

Sparsely connected autoassociative lattice memories (SCALMs) are very general models defined on complete lattices, a mathematical structure which is obtained by imposing some ordering on a set. They are computationally cheaper and mathematically simpler than "traditional" models and other memories such as the original autoassociative morphological memories (AMMs) of Ritter and Sussner because they only compute maximums and minimums. This paper provides theoretical results on SCALMs defined on a general complete lattice as well as an application of these memories for the storage and recall of color images. Precisely, we characterize the recall phase of SCALMs in terms of their fixed points. Then, we show that any endomorphic lattice polynomial-a concept that generalizes the notion of lattice polynomial of Birkhoff-on the fundamental memory set represents a fixed point of the SCALMs. Also, we discuss the relationship between SCALMs and the original AMMs. Finally, we provide some experimental results on the performance of SCALMs, defined on different color lattices, for the reconstruction of color images corrupted by either Gaussian or impulsive noise.

• 出版日期2012-11