Sentence examples for bpfa with from inspiring English sources
We also see that BPFA with the sparse inpainting scheme improves over the original BPFA inpainting procedure for 30 and 50% corruption.
Consistency: Figures 5 and 6 show the dictionaries and if applicable low-rank components for L=1 learned by ITKrM M), (w KSVD and BPFA with SVD initialisation from uncorrupted and corrupted data.
Similarly, the BPFA dictionary with SVD initialisation performs much better than the randomly initialised one.
In all cases, we show that our results are mostly better than the ones of BPFA and wKSVD, with a large reduction of the computational costs with respect to the latter.
For KSVD with L=1 and BPFA, we use the same setup as described above.
In the first experiment, we compare the ITKrMM dictionaries to those learned with wKSVD and BPFA.
In particular, we compare to the inpainting schemes based on wKSVD and BPFA dictionaries as well as analytic dictionaries such as the DCT basis and the overcomplete DCT frame with 144 atoms.
The best result for each setting is marked in bold Table 5 Comparison of the SSIM value (0.−) for inpainting of images with various corruption levels based on analytic dictionaries, DCT and MCA, and dictionaries learned on all available corrupted image patches, BPFA, BPFAomp, wKSVD and ITKrMM (modified inpainting is marked with a *) Algorithm Bar.
Compared with conventional dictionary learning methods, the introduced dictionary learning method is based on a popular beta process factor analysis (BPFA) for an adaptive learning performance.
Fig. 8 Approximation quality of dictionaries with low-rank components of various sizes on Barbara, DCT144 as well as BPFA, (w KSVD and ITKrM M) learned on uncorrupted data (left) and on 50% corrupted data (right).
