dncnns with the matlab functions denoiseimage and denoisingnetwork  (MathWorks Inc)

Journal: Scientific Reports

Article Title: A multi-parameter persistence framework for mathematical morphology

doi: 10.1038/s41598-022-09464-7

Figure Lengend Snippet: Conceptual images with salt and pepper noise and the results of different denoised algorithms: AMF , denoiseImage, NAMF , Algorithm 1, and CNN with median layers . The parameters for Algorithm 1 are MaxIter=10 and Sizetol=5. Figure a is the ground truth image of the experiment.

Article Snippet: Some state-of-art denoised algorithms are also considered in the experiment, such as the denoiseImage function in Matlab , adaptive median filter (AMF) , non-local adaptive mean filter (NAMF) , and a convolutional neural network (CNN) with median layers .

Techniques:

Journal: Scientific Reports

Article Title: A multi-parameter persistence framework for mathematical morphology

doi: 10.1038/s41598-022-09464-7

Figure Lengend Snippet: The ground truth images are ( a ), ( c ) where ( a ) is one of images from the open domain database Flickr Material Database and the full database can be accessed at https://people.csail.mit.edu/celiu/CVPR2010/FMD/ . ( c ) Adopted from the Wikipedia page https://en.wikipedia.org/wiki/File:Mona_Lisa.jpg . The unprocessed images with noise are ( b ) and ( d ). The first two rows are the denoised images obtained by the following algorithms: AMF , denoiseImage, NAMF , Algorithm 2, 3, and CNN with median layers . The last row is the averaged IOU, PSNR, and SSIM scores for the noised grayscale image for ( a ) and by color channel for ( c ).

Article Snippet: Some state-of-art denoised algorithms are also considered in the experiment, such as the denoiseImage function in Matlab , adaptive median filter (AMF) , non-local adaptive mean filter (NAMF) , and a convolutional neural network (CNN) with median layers .

Techniques:

Journal: Scientific Reports

Article Title: A multi-parameter persistence framework for mathematical morphology

doi: 10.1038/s41598-022-09464-7

Figure Lengend Snippet: Conceptual images with salt and pepper noise and the results of different denoised algorithms: AMF , denoiseImage, NAMF , Algorithm 1, and CNN with median layers . The parameters for Algorithm 1 are MaxIter=10 and Sizetol=5. Figure a is the ground truth image of the experiment.

Article Snippet: To compare, we consider the following algorithms in the literature: the denoiseImage function provided by Matlab 2019, adaptive median filter (AMF) , non-local adaptive mean filter (NAMF) , and a convolutional neural network (CNN) with median layers .

Techniques:

Journal: Scientific Reports

Article Title: A multi-parameter persistence framework for mathematical morphology

doi: 10.1038/s41598-022-09464-7

Figure Lengend Snippet: The ground truth images are ( a ), ( c ) where ( a ) is one of images from the open domain database Flickr Material Database and the full database can be accessed at https://people.csail.mit.edu/celiu/CVPR2010/FMD/ . ( c ) Adopted from the Wikipedia page https://en.wikipedia.org/wiki/File:Mona_Lisa.jpg . The unprocessed images with noise are ( b ) and ( d ). The first two rows are the denoised images obtained by the following algorithms: AMF , denoiseImage, NAMF , Algorithm 2, 3, and CNN with median layers . The last row is the averaged IOU, PSNR, and SSIM scores for the noised grayscale image for ( a ) and by color channel for ( c ).

Article Snippet: To compare, we consider the following algorithms in the literature: the denoiseImage function provided by Matlab 2019, adaptive median filter (AMF) , non-local adaptive mean filter (NAMF) , and a convolutional neural network (CNN) with median layers .

Techniques: