Using Color Classification for Logos Retrieval

Chiun-Hsiun Lin

《資訊電子學刊》 8卷2期 (2019/11) Pp. 21-28

摘要

A color classification for logos retrieval scheme is developed to retrieve logos efficiently. We transfer each logo image to only 8 colors image using only 8 rules. Afterwards, we utilize the 8 colors images for comparing the logos. The major differences between our proposed method and existing ones are: 1. We consider the spatial information. (Our approach is permutation comparison). Most of the existing ones (color classification-based methods) are combination comparison. 2. We resized all images to 20 X 20 pixels. Therefore, our approach can handle diverse sizes, defocus and noise problems. 3. We adopt the concept of the relative relation of RGB. Consequently, our system can deal with dissimilar lighting conditions, partial occlusion, and various color saturation at the same time. 4. We don't want to know what the meanings of the logos. We really want to know what the color and color location of the logos. Since the 8 colors comparison is very fast, our approach is very speedy. Our approach also makes logos searching simple and robust.

關鍵字

Content-based images retrieval (CBIR) ； Color classification ； Logos retrieval