Hash based manifold learning technique to generating random fields for image segmentation

Abstract

The manifold learning technique is a class of machine learning techniques that converts the intrinsic geometry of the data from higher to lower dimensional representation by using the manifold distance and preserved in hamming space. It is an offline learning process, so it requires more time and memory. We proposed a new hash based manifold learning technique to generate random fields for image segmentation. The proposed method is a two-step process, first is to find the similarity neighborhood pixels, the second is to construct the weighted matrix. The proposed method requires less time and space comparatively than the existing method, because the manifold structure is directly reconstructed in hamming space, which has been shown in the experimental results on semantic datasets and our own datasets in the form of tables and graphs.