Feature extraction method of halftone images based on pixel aggregation descriptor

摘要

Since halftone images only have the values of 0 and 1, the amount of information that can be used in the feature extraction process is relatively little. The existed feature extraction methods have some disadvantages such as too many dimensions, longer extraction time, and lower classification accuracy. In view of the above questions, a pixel aggregation descriptor is proposed to extract the feature of halftone images in this paper. Set operations are introduced to obtain this pixel aggregation descriptor, which makes the texture of the images clearly visible and greatly reduces the dimension of feature and the time consumption of classification. Then, the correlations between pixels, including distance and angle are taken as factors to improve the accuracy in image classification. In experiments, the reliability and stability of the feature extraction methods are evaluated and compared. Then, the influence of parameters, time complexity, and the limitations are discussed. Experiments show that our method realized the effect of higher speed and precision in feature extraction.

关键词