Abstract

It is common knowledge that local color histograms are a straightforward and often used method for doing color-based picture retrieval. But there is a significant amount of space above it. A grid of cells that is overlaid on the photographs and the use of histograms for each color both contribute to a reduction in the amount of storage space required. The technique in question thus becomes sensitive to the rotation and translation of images. The purpose of this study is to provide a unique method for modeling picture similarity that we term HG. This method makes use of a weighted undirected graph for each color and each vertex of the graph represents a bin of the histogram distribution. As a consequence of this the method is able to take use of color location and need less space overhead but it is not responsive to rotation or translation. We decided to conduct an experiment using a library of photographs that had 9600 different landscape pictures. The results of our experiments indic