Computational Imaging Group

TAMPERE17 noise-free image database


300 color images 512x512 pixels without noise (variance < 1), without interpolation and without lossy compression.

Color images: tampere17_color.zip (140 Mb).
Standard deviations of noise for each color image: std_color.txt.
Grayscale versions of images: tampere17_grayscale.zip (46 Mb).
Standard deviations of noise for each grayscale image: std_grayscale.txt.

Please refer to the following paper:
M.Ponomarenko, N.Gapon, V.Voronin, K. Egiazarian, "Blind estimation of white Gaussian noise variance in highly textured images", Image Processing: Algorithms and Systems XVI, 2018, 5p. (pdf) (code)

Short description:
The TAMPERE17 database has 300 test images of 512x512 pixels (both color and grayscale versions). Images for the database are obtained with monitoring of noise level (the photo shooting was carried out on calibrated cameras with fixed ISO) and without any interpolation, lossy compression or other spatial processing (denoising, sharpening, etc.). For each image, a fragment 3072x3072 pixels or 2560x2560 pixels (depending on digital camera) was selected, and downscaled to 512x512 (by averaging of pixel values for regions 6x6 pixels or 5x5 pixels respectively). As a result, only images with standard deviation of noise less than 1 have been included in the database. For the grayscale version of images, the noise level is even lower (values of noise standard deviation are in the text file included into the database).

Examples of images: