问题
I am using LBP with MATLAB for extraction feature but the accuracy is too low
how to reduce the feature bins in LBP?
many thanks.
回答1:
Use the pcares function to do that. pcares stands for PCA Residuals:
[residuals, reconstructed] = pcares(X, ndim);
residuals returns the residuals obtained by retaining ndim principal components of the n-by-p matrix X. X is the data matrix, or the matrix that contains your data. Rows of X correspond to observations and columns are the variables. ndim is a scalar and must be less than or equal to p. residuals is a matrix of the same size as X.
reconstructed will have the reduced dimensional data based on the ndim input. Note that reconstructed will still be in the original dimension as X. As such, you can choose the first ndim columns and this will correspond to those features constructed using the number of the dimensions for the feature specified by ndim. In other words:
reduced = reconstructed(:,1:ndim);
As such, reduced will contain your data that was dimension reduced down to ndim dimensions.
Small Note
You need the Statistics Toolbox in order to run pcares. If you don't, then this method won't work.
来源：https://stackoverflow.com/questions/24805898/how-to-improve-lbp-operator-by-reducing-feature-dimension