Video Corner-Logo Recognition System And Matching Algorithm

Abstract

Video corner-logo is widely used in television programs and online videos to provide video column information or advertising information in a variety of forms. Video corner-logo detection and recognition algorithm study is therefore getting more and more attention due to the practical needs. Effective detection and recognition algorithm is the key to corner-logo extraction and corner-logo information analyzation. In this paper, a whole framework and design flow of video corner-logo detection and recognition system are introduced. The design of video corner-logo recognition algorithm is discussed in details and analyzed. The special characteristics of corner-logo makes it difficult to detect, extract and recognize corner-logo. Aiming at this problem, this paper proposes an effective corner-logo recognition algorithm, SURF based color invariance and shape context algorithm. The algorithm calculates the color invariance information of the image, constructs the color descriptor, extracts the interest points of the image, at the same time, introduces shape feature descriptor of the shape context, combines the two kinds of feature descriptions into a new joint descriptor with optimal weight coefficient, and then the similarity is calculated by means of Euclidean distance to match the interest points. The experimental results show that this recognition system and matching algorithm can improve the image recognition accuracy and can recognize corner-logo effectively.