A Novel Logo Detection and Recognition Framework for Separated Part Logos in Document Images
Sina Hassanzadeh, Hossein Pourghassem · 2011
The procedure of detection and recognition for separated part logos in document images is a major problem in logo detection and recognition algorithms that previously have presented. In this paper, a novel logo detection and recognition framework based on spatial and structural features especially for separated part logos application is proposed. To overcome this problem, we consider some specifications of these logos such as centroid coordinates and intersection of each logo's separated part bounding box in detection process. To improve detection performance, a morphological dilation operation is used to merge separated parts of logos. In our framework, a new spatial feature for logo recognition is presented. This feature is defined based on histogram of object occurrence in a new tessellation of logo image. KNN is used to recognize the detected logos. Our proposed framework is evaluated on a public document image database for detection process and standard logo dataset of Maryland University for recognition process. The provided results show its performance in logo detection and recognition.