Robust Picture Search by Absent Color Indexing

Picture search has become a vital part of image processing where color histogram-based methods can effectively identify objects. However, conventional color histogram-based methods are not sensitive to similar objects. This paper aims to propose a novel scheme for robust picture search based on Absent Color Indexing, ABC in short. It generates two-dimensional color histograms via CIE Lab color space. Then, we utilize a threshold \( h_{T} \) to divide them into an apparent color histogram and another absent color histogram. To enhance the performance by absent colors, we invert absent color histograms. Finally, we calculate similarity by histogram intersection. Experimental results on Mondrian random patterns and real-scene pictures show that the proposed ABC method is more robustness than conventional color histogram-based methods. Especially for similar searching objects, it has a large margin to distinguish different objects.