Sparse filtering for image classification applications

Let us consider a feature distribution matrix over a finite dataset, where each row is a feature, and each column is an example. I have tried the code in the link below to get sparse features in order to use them for image classification. In the training stage, I have put the features of each of the training images in one of the columns of the matrix, and got the sparse features needed for training the classifier. My question is: How can I get the sparse features for the test images? How can I apply sparse filtering for only one image? The algorithm depends on putting all test images on one matrix; a column for each image. The algorithm depends on some normalization operations applied on columns and vectors. In real situations, it is not reasonable for the test images to depend on each other, and we may need to test only one image.

https://www.mathworks.com/help/stats/workflow-for-feature-extraction.html?fbclid=IwAR35BqYe87lNJWTaUOGdqTrDYjTiVJ-s1TnasJNVlbZxCm4zlUoCiSIWgWo

Thanks