IPOL Journal · The Noise Clinic: a Blind Image Denoising Algorithm

The Noise Clinic: a Blind Image Denoising Algorithm

Please cite the reference article if you publish results obtained with this online demo.

Noise Clinic Automatic Noise Estimation and Denoising.

The test images are divided into three groups:

Raw images downscaled by 2 to have one raw value (R, G, B) at each pixel. Their raw values are multiplied by 32 in order to be able to visualize them. In the list, they are are refered to as "raw".
Many noisy images without a known noise model. They might have suffered all kind of alterations.
High SNR raw images, downscaled by 8 and the color channels averaged, so they are nearly noiseless. In the list, they are refered to as "no noise".

Select Data

Click on an image to use it as the algorithm input.

Dog

Frog

Girl

Circuit

Bears

Chupa Chups

Fibers

Lena

Movie 1

Objects

Pentagon

Room

Ship

Singer

Synthetic

Vehicle

Very Noisy

Vinegar

Woman

no noise: Computer

no noise: Dice

no noise: Traffic

Girls

Pattern

Pattern 2

Cells

Elvis

Family

Postcard

Solvay Conf., 1911

raw: ISO 1600, t=1/30

raw: ISO 1250, t=1/30

raw: ISO 1600, t=1/30

Marilyn1

Marilyn2

Koala

Museum

Movie 2

Cards

Beer

Solvay Conf., 1913

Solvay Conf., 1922

Solvay Conf., 1927

Old postcard 1

Old postcard 2

Old postcard 3

Old postcard 4

Old postcard 5

Old postcard 6

Old postcard 7

Old postcard 8

Old postcard 9

Old postcard 10

Old postcard 11

Old postcard 12

Old postcard 13

Infrared

Palace

Moire

Night Vision

image credits

Select Data

Upload Data