首页>
外国专利>
Nonlocal mean image denoising with detail preservation using blends driven by self-similarity
Nonlocal mean image denoising with detail preservation using blends driven by self-similarity
展开▼
机译:使用由自相似驱动的混合来保留细节的非局部均值图像去噪
展开▼
页面导航
摘要
著录项
相似文献
摘要
A system, apparatus, method, and computer-readable medium for texture enhanced non-local mean (NLM) image denoising. In embodiments, the detail is maintained in the filtered image data through a blend between the noisy input target pixel value and the NLM pixel value driven by self-similarity, and an independent local texture. Further notification by the scale of. In embodiments, the blend is driven by one or more blend weights or coefficients indicative of the texture, whereby the level of detail retained by the enhanced noise reduction filter is scaled by the amount of texture. Embodiments herein may thereby more aggressively denoise areas of the image that are significantly lacking texture (ie, smooth) than coarser texture areas. In some further embodiments, the blending factor is further determined based on the similarity score of the candidate patches, and the number of these scores considered is based on the texture score.
展开▼