Automatic image enhancement


Following code is useful for automatically enhancing under and over exposed images.

Automatic contrast enhancement using reversible data hiding (acerdh):

     Code: [github code], [gitlab code], [gitee code], [Cite]
     Paper: [Accepted version], [Official version]

Following image dataset is useful for testing automatic image enhancement techniques to make under or over exposed images more visble to human eyes.

Under and over exposed image dataset:

     [Color PNG], [Grayscale PNG], [Cite]

SNS image compression


Following dataset is useful for testing new image compression techniques that are specifically targeting SNS platform. Currently, we offer images that are uploaded and downloaded from Instagram SNS. We hope to expand this work in the future.

Instagram compression test dataset: This contains resized and unenhanced jpeg images. All images are resized to an optimal size for Instagram post (1080p x 566p ~ 1350p) before uploading it to Instagram. Since an Instagram post can have multiple number of images, dataset is organized by number of images in a post. Each dataset contains the uploaded original version, downloaded version, and an image dataset description file.

     1 image per post (20 posts):
      [Color JPEG], [Grayscale JPEG], [Color HEIC], [Cite]
     2 images per post (10 posts):
      [Color JPEG], [Grayscale JPEG], [Cite]
     3 images per post (10 posts):
      [Color JPEG], [Grayscale JPEG], [Cite]
     4 images per post (10 posts):
      [Color JPEG], [Grayscale JPEG], [Cite]
     5 images per post (10 posts):
      [Color JPEG], [Grayscale JPEG], [Cite]
     6 images per post (10 posts):
      [Color JPEG], [Grayscale JPEG], [Cite]
     All Instagram compression datasets:
      [Color JPEG], [Grayscale JPEG], [Cite]

SNS built-in enhancement


Following dataset is useful for testing the effect of built-in SNS image enhancement.

Instagram filter dataset: This dataset contains resized and filter-applied JPEG images. All images are resized to an optimal size for Instagram post (1080p x 566p ~ 1350p) before uploading it to Instagram. The dataset contains images enhanced with 41 image filters (1977, Aden, Amar, Ashb, Bran, Broo, Char, Clar, Crem, Dogp, Earl, Ging, Ginz, Hefe, Hele, Huds, Inkw, Juno, Kelv, Lark, Lo-F, Ludw, Mave, Mayf, Moon, Nash, nfit, Perp, Reye, Rise, Sier, Skyl, Slum, Stin, Sutr, Toas, Vale, Vesp, Wald, Will, X-Pr). Since an Instagram post can have multiple number of images, dataset is organized by number of images in a post. Each dataset contains the uploaded original version, downloaded version, and an image dataset description file.

     1 image per post (20 posts for each filter):
      [Color JPEG], [Grayscale JPEG], [Cite]
     2 images per post (10 posts for each filter):
      [Color JPEG], [Grayscale JPEG], [Cite]
     3 images per post (10 posts for each filter):
      [Color JPEG], [Grayscale JPEG], [Cite]
     6 images per post (10 posts for each filter):
      [Color JPEG], [Grayscale JPEG], [Cite]
     All Instagram filter datasets:
      [Color JPEG], [Grayscale JPEG], [Cite]