Automatic Logo Recognition System from The Complex Document Using Shape and Moment Invariant Features

One of the strongest clues for retrieval of content information from complex document images is logo. However, due to the wide range of transformations that an object might undergo, this is also the most difficult one to handle. Logo retrieval is one of the major barriers now a days for image databases being commonly used. Automatic logo detection and recognition continues to be of great interest to the document retrieval community as it enables effective identification of the source of a document. In this paper, a new approach is proposed for detection of logo and extraction from the document images that robustly classifies and precisely localizes logos using a boosting strategy across multiple image scales. At a coarse scale, logo recognition system also comprises of three phases: Preprocessing, feature extraction and features matching. For feature extraction 10 shape features and 07 hu’s moment invariant features have been adopted. The Euclidian Distance (ED) is taken as a similarity measure parameter for the features matching with Nearest Neighbor, K Nearest Neighbor and Support Vector Machine Classifier are also compared. The accuracy of 96.24% from the SVM classier proved to be potential classifier from the experimental results as compared to 88.21% for NN and 91.46% from KNN classifiers.