Relaxed Block-Diagonal Dictionary Pair Learning With Locality Constraint for Image Recognition

Fusion of motion smoothing algorithm and motion segmentation algorithm for human animation generation

In the field of human animation generation, the existing technology is often limited by the dependence on large-scale data sets, and it is difficult to capture subtle dynamic changes when processing motion transitions, resulting in insufficient animation fluency and realism. In order to improve the naturalness and diversity of human animation generation, a method combining motion smoothing algorithm and motion segmentation algorithm is proposed. Firstly, the tree-level model based on human skeleton topology and bidirectional unbiased Kalman filter are used for noise reduction pre-processing of motion data to improve the accuracy of motion capture. Then, combining the discriminant analysis algorithm based on sparse reconstruction and the multi-scale temporal association segmentation algorithm, the key motion segments of the behavior pattern change are identified adaptively. The experimental results show that the accuracy of the proposed algorithm reaches 0.96 in coarse-grained segmentation and 0.91 in fine-grained segmentation, and the segmentation time is 15 seconds on average, which significantly exceeds the prior art. In addition, the algorithm shows superior results in color fidelity, detail representation, motion fluency, frame-to-frame coherence, overall animation consistency, action authenticity, and character expressiveness, and the average user satisfaction is above 0.85. The research not only enhances the naturalness and diversity of human body animation, but also provides a new impetus for technological advances in computer graphics, virtual reality and augmented reality.

Discriminative analysis dictionary learning with adaptively ordinal locality preserving

Dictionary learning has found broad applications in signal and image processing. By adding constraints to the traditional dictionary learning model, dictionaries with discriminative capability can be obtained which can deal with the task of image classification. The Discriminative Convolutional Analysis Dictionary Learning (DCADL) algorithm proposed recently has achieved promising results with low computational complexity. However, DCADL is still limited in classification performance because of the lack of constraints on dictionary structures. To solve this problem, this study introduces an adaptively ordinal locality preserving (AOLP) term to the original model of DCADL to further improve the classification performance. With the AOLP term, the distance ranking in the neighborhood of each atom can be preserved, which can improve the discrimination of coding coefficients. In addition, a linear classifier for the classification of coding coefficients is trained along with the dictionary. A new method is designed specifically to solve the optimization problem corresponding to the proposed model. Experiments are performed on several commonly used datasets to show the promising results of the proposed algorithm in classification performance and computational efficiency.