Survey on Image Retrieval Based on Rotation, Translation and Scaling Invariant Features

Abstract

How to retrieve information from the massive image database in time is a “bottleneck” faced by network information processing in recent years and has become a research hotspot at home and abroad. For unstructured image data, we have studied the traditional text-based retrieval method is inefficient, and a content-based image retrieval technique. Compared with standard color, texture and shape features, the image features with the same scale of rotation and translation can get better retrieval results. Therefore, image retrieval based on invariant features has broad research prospects. In this paper, the algorithm of image retrieval based on invariant features is studied. Combined with the principle of integral invariant construction of geometric transformation group and the extraction principle of scale-invariant feature points, two new rotation, translation, and scale-invariant features are presented. That is, the rotation, scaling and translation (RST) invariant feature extraction method and these features are applied to image retrieval.