Method of Clarity Evaluation of Shoes in the Video Based on Multi-feature Fusion

Regarding the difficulty in analyzing shoes with varying clarity in videos, a multi-feature fusion based clarity evaluation method is proposed and studied. First, shoe images are detected and cropped from videos and then operators including HSV, GLCM and DWT are used to extract the color and texture features. Then the extracted features are fused with features from deep convolutional network to form the clarity score of the corresponding shoe image. Experiments show that the output of the proposed method is highly consistent with human subjective feelings. Thus, the proposed method can be used as a data filtering preprocess method to rule out vague shoe images, offering efficient and reliable digital support for subsequent shoe sample identification.