An Ensemble Approach to Graph Layout for Movie Poster Similarity Visualization
Cao Vu Bui, Thi-Thu-Hong Phan, Luong Vuong Nguyen · 2023
An integral element within the realm of movie-making is the movie poster image. This visual medium holds significant importance, conveying essential details about the movie, including its title, characters, and genre. Hence, the features extracted from the movie poster are valuable for measuring the similarity of the movies. This study proposes an ensemble approach to identify movie similarity for visualizing the similarity of movie posters. Specifically, we consider three methods i) reference-based, ii) traditional feature-based, and iii) deep learning feature-based regarding constructing the similarity matrix. This way, we can estimate the movie similarity more accurately to provide it as the input dataset for personalized movie recommendation systems. In particular, our method focuses on extending the color-based search algorithms that use low-level image histogram features. This approach helps the classification processing time increase, leading to the quick generation and comparison of feature vectors.