컴퓨터 알고리즘 추천에 의한 포토모자이크 기법

Abstract

We propose a new photomosaics method using principle components of candidate photo set. PCA(Principle Component Analysis) reduces dimensions of non-intuitive high dimension data set and allows us to deal the data set more easily. Principle component is a good direction of a basis representing each data according to its distribution. Two data projected onto a certain principal component are similar if their projected position are similar. More principle components provide more accurate comparison between data. We selected the most similar candidate photo to an input photo tile using properties of PCA. Unlike most previous photomosaics methods,photomosaics author can choose candidate photo set and manipulate the number of principle components to change mood of the result of his work. For more active making of photomoasics, we expend room for intervention of author's opinions and propose a method to make various results by a simple way.