Trademark image retrieval based-color features using statistical methods

Abstract EN

Since the number of registered trademarks are increasing rapidly, manage these images and identifying similar trademarks by human in response to user queries has become a significant problem and time-consuming.

To deal with this problem, design and implement an effective queried by Image Retrieval System (IRS) called Trademark Image Retrieval Based-Color Feature (TIRBCF) which is able to retrieve trademark image based on visual low-level image content from a big database of trademark images described by descriptors of color feature is presented as the major objective of this work.

TIRBCF is based over the index of Hue (H), Saturation (S) and Value (V) of HSV color space because it is so close to human visual vision.

TIRBCF is automatically selects appropriate color descriptor using statistical method to discriminate trademark images.

Color retrieval is achieved by utilizing four statistical color descriptors including HSV Color Histogram (HCH), Mean HSV Color, and Median HSV Color (MHC) of an image as a feature set.

Intersection Distance (ID) measurement for HCH and Euclidean Distance (ED) for other color descriptors are used as the color vector matching because its simplicity and effectiveness.

The user provides query image through user interface, the system will extract color descriptors from it, then compared with those extracted from the trademark images database in order to retrieve similar trademarks.

Experiments have been conducted on a database of 3000 trademark images and the results demonstrate the accuracy of the precision/recall estimates in comparison to human-judge.