dcscn-super-resolution icon indicating copy to clipboard operation
dcscn-super-resolution copied to clipboard

some questions about images......

Open wehkjhkl opened this issue 5 years ago • 1 comments

Are most existing SR methods suitable for images collected from surveillance video?the quality of images are so poor,the image format is .jpg .I run some code but the results show that it is useless.

wehkjhkl avatar Jun 14 '19 01:06 wehkjhkl

Hi, usual super resolution model won't work for motion jpeg or video images. This is because usually SR model is trained by normal (not jpeg or video) images. You need to train with those compressed images to restore high resolution images for that case.

jiny2001 avatar Jun 18 '19 06:06 jiny2001