1：Experimental Design and Method Selection:

The method combines four existing criteria with a new feature computed by a convolutional neural network to predict channel-mismatch probability. A logistic-regression model is trained on these features for final prediction.

2：Sample Selection and Data Sources:

A training data set of 1,000 stereoscopic scenes from various movies, half with artificially swapped views, and a test set of 900 scenes from other movies.

3：List of Experimental Equipment and Materials:

Fast local block matching for disparity maps and optical flow, fast global smoothing filter for refining disparity maps and optical-flow fields.

4：Experimental Procedures and Operational Workflow:

Calculation of feature values for each frame, smoothing outliers, training logistic-regression model, and testing on the prepared data set.

5：Data Analysis Methods:

Evaluation using accuracy, area under ROC-curve (AUC), and F-measure metrics.