GIF Image Retrieval in Cloud Computing Environment

Abstract

GIF images have been used in the last years, especially on social media. Here it is explored a content-based image retrieval system to work specifically with GIF file format. Its implementation is extended to a cloud computing environment. Given the Tumblr GIF dataset, it is created a “search by example” image retrieval system. To describe the images, low-level features are used: (1) color, (2) texture and (3) shape. The system performs the search using just GIF images as query images. To obtain faster results on the retrieval process, a hashing indexing approach is used. The system showed a complexity of O(n2) for indexing and O(log(n)) for retrieval. Additionally, better results were obtained (in relation to precision and recall) for simple images, instead of images with a lot of movements.