Eigenmosaic: Reconstruction of Images Using PCA Based Mosaic Method

In this paper we have introduced new techniques for generating compound images, or mosaics, which include reconstruction of main images using homomorphous square patches. These patches are extracted from another image, called the Key Image, and are placed next to each other in order to produce the mosaic image. To perform that, criterions like color average, variance, and histograms have been used so far due to their important role in describing the facade of an image. PCA is a valuable statistical method in recognition of pattern among the data; that can be used on images as well. These methods basically help us to separate and analyze the image parameters. PCA implementation results will be used as filters, eigen filters, for the main image leading to extraction of constructor lines and different parts of image construction texture. In the proposed approach, the extracted parameters from the main and key image can be a local or global reference to select parts in order to produce the mosaic. The results of experiments have shown that our PCA-based method performance well in building mosaics.