A binary neural shape matcher using Johnson Counters and chain codes

Abstract

Images may be matched as whole images or using shape matching. Shape matching requires: identifying edges in the image, finding shapes using the edges and representing the shapes using a suitable metric. A Laplacian edge detector is simple and efficient for identifying the edges of shapes. Chain codes describe shapes using sequences of numbers and may be matched simply, accurately and flexibly. We couple this with the efficiency of a binary associative-memory neural network. We demonstrate shape matching using the neural network to index and match chain codes where the chain code elements are represented by Johnson codes

Edge Tracing Based on Improved Genetic Algorithm

“…However, it is not desirable to trace chain code for complicated and noisy edges because of considerable computational complexity. Existing studies about chain code are focused on areas such as image compression [12][13][14], shape analysis [15][16], and object matching [17][18], and most of them are based on the assumption that we have obtained desirable chain code of the edge.…”