usage:

1.saliency-based sampling
    1) open LLC_Test_from_txt.m
	2) set parameter
	3) run

2.random sampling
    1) open LLC_Test.m
	2) set parameter
	3) run
	
3.Harris sampling
    1) open LLC_Test_Harris.m
	2) set parameter
	3) run



folder:
    dictionary: random samples and dictionary
	image: training and testing images
	Liblinear: linear SVM matlab
	points_fix: interest points sampled by Saliency-based sampling
	sift: sift toolbox
files:
    Calculate4DirectionSobelPrewittDescriptor.m: random sampling and SPF descriptors
	Calculate4DirectionSobelPrewittDescriptorFromTXT.m: Saliency-based sampling and SPF descriptors
	Calculate4DirectionSobelPrewittDescriptorHarris.m: Harris and SPF descriptors
	CalculateSobelFeatures.m: function for generate histogram feature for each patch
	CalSobelDesc_save.m: save function to use in 'parfor'. matlab save function cannot be used in 'parfor'
	DictionaryFormulation.m: sampling SPF descriptors to generate codebooks by vl_kmeans
	extr_sift: call function to extract sift descriptors
	extr_sobel_prewitt_4_direction.m: 
	extr_sobel_prewitt_4_direction_harris.m:
	extr_sobel_prewitt_from_txt.m:
	LLC_coding_appr.m: original function download from LLC
	LLC_extr_par_save.m: save function to use in 'parfor'. matlab save function cannot be used in 'parfor'
	LLC_pooling.m: original function download from LLC
	LLC_Test.m: main function to random sampling-> SPF descriptors->sparse representation->linear SVM->recognition rate
	LLC_Test_all_multiscale_con.m: main function to MSR->linear SVM->recognition rate
	LLC_Test_from_txt.m: main function to saliency-based sampling-> SPF descriptors->sparse representation->linear SVM->recognition rate
	LLC_Test_Harris.m: main function to harris sampling-> SPF descriptors->sparse representation->linear SVM->recognition rate
	multiscale_concatenate.m: main function to concatenate SSR->MSR