Multiscale Hybrid Nonlocal Means Filtering Using Modified Similarity Measure

Figure 4

The flow chart of the proposed MHNLM algorithm. The given image is decomposed up to two scale levels using DSWT and accelerated nonlocal means [34] is applied on the detail data (HL, LH, and HH) at each scale. Afterwards, the predenoised image is obtained using IDSWT. Finally, the modified similarity measure based nonlocal means filtering is employed on the given noisy image using reference normal vector patches extracted from the predenoised image.