BPFA.zip_BPFA

File List: Mingyuan_nips2009_final.pdf: the final version of the paper "Non-Parametric Bayesian Dictionary Learning for Sparse Image Representations," Neural Information Processing Systems (NIPS), 2009. Model and Inference.pdf: The model and inference for the above paper. house.png: the orignial house image. barbara.png: the original barbara image. castle_damaged.png: the castle image with 80% pixels missing. castle_original.png: the orignial castle image. fig13_damaged.png, fig13_damaged.bmp: the New Orleans image corrupted by text. fig13_original.png: the original New Orleans image. Solve out of memory error.txt: Instructions for allowing Matlab to manage 3GB memory instead of the default 2GB memory in Windows XP. Demo files: Demo_Denoising.m: Running file for BPFA image denoising. Demo_Inpainting_GRAY.m: Running file for BPFA gray scale image inpainting. Demo_Inpainting_RGB.m: Running file for BPFA rgb image inpainting. Main programs: BPFA_Denoising.m: The BPFA grayscale image denoising program. BPFA_Inpainting.m: The BPFA image inpainting program. Subprograms for Gibbs sampling: SampleDZS.m: Sampling the dictionary D, the binary indicating matrix Z, and the pseudo weight matrix S. Used for no missing data case. SampleDSZ_MissingData: Sampling the dictionary D, the binary indicating matrix Z, and the pseudo weight matrix S. Used when there are missing data. SamplePi.m: Sampling Pi. Samplephi.m: Sampling the noise precision phi. Samplealpha: Sampling alpha, the precision of si. Subprograms for the updates in sequential learning XUpdate.m: Update the training data set in sequential learning. Used for no missing data case. XUpdate_MissingData.m: Update the training data set in sequential learning. Used when there are missing data. SZUpdate.m: Update the pseudo weight matrix S and binary indicating matrix Z in sequential learning. idexUpdate.m: Update the index of the training data in the input data matrix X. YfalgUpdate.m: Update the binary mask matrix Yflag associated with the input data matrix X. Other subprograms: sparsity.m: Squeeze out zero components in the sparse matrix. DispDictionary.m: Display the dictionary elements as a image DenoiseOutput.m: Reconstruct the image InitMatrix.m: Initialization