PhotoMagic - [PKU Hackers]

 

Project Design Motivation

Nowadays, the mobile image processing apps, such as Instagram, mainly do photo toning from a global view. However, for most of the landscape photos, whose brightness usually covers a large span, this method is naturally flawed. Meanwhile, as to PC software, for example, Photoshop; despite of its powerful functions, selection is always indispensable and toning is as a result painful and time-taking.   By accepting the ideas of image partition and semantics analysis, we have achieved a powerful automatic image processing tool. It can automatically recognize different objects and then make adjustment according to its properties, for example, recognizing different facades of walls and fixing backlight shadow problems. Because of its outstanding performance and portability, the project can greatly benefit the common people who have little knowledge about photo toning. Meanwhile, for photographers, who always have huge quantities of landscape photos in their camera, the product can also save plenty of time. Besides, considering the transplantable developing tools used, the product can be easily implemented on any mobile or embedded system.  

Technical Issues

The project mainly exploits image partition and semantics analysis technologies, and develops series of image processing algorithms to implement auto toning. As a preprocess for photo toning, image partition is used for making selections from the photo. Down-sampling and k-means partition are utilized to get a proper selection without too many noises. The project used Wolfram as an approach to semantics analysis. With semantic explanation of photos, Photomagic decides how to tone the photo based on common sense like the sky is blue. Many basic image processing algorithms are also implemented to tone the photo, for example color balance is exploited for color adjustment and Gaussian blur is applied in glow effects.    

Creators

 Hackers from Peking University

Share this project:

Updates