A Feature Encoding Based on Low Space Complexity Codebook Called Fuzzy Codebook for Image Recognition

Abstract

For image recognition, a codebook approach is generally used for representing an image to a feature vector. In this approach, the codebook, which is visual vocabulary, is generated for each local feature framework in the case of using multiple local feature frameworks. Hence, the codebook is required that is a small memory footprint. Image representation based on a compact codebook by using fuzzy clustering has been presented, but it includes a high computational complexity operation. This paper presents a reducing computational complexity in the image representation step and experimental results of online classifier.