Image matting

Automatic blurred non-blurred regions identification and image matting
Total votes: 3

Degree:

Automatic blurred non-blurred regions identification and image matting

In the first part of this project we have implemented three different techniques in order to identify blurred and non-blurred regions in an image.

At first we show the results for each single method, stressing about pros and cons of each of them. Finally we combine all of three techniques in order to obtain an unique more effcient algortihm.

In the second part we exploited work done previously. In particular starting from 'maps' containing informatin about blurred and non-blurred regions, we can obtain an opacity map of the image.

In the project we show both illustrated results (final maps) and graphics representing analysis of the test (we exhibit recall-precision-accuracy evolution).

At the end we mention about some possible future development afford to enhance the results.

References: 

[1] "Image Partial Blur Detection and Classication". Renting Liu, Zhaorong Li, Jiaya Jia
[2] "An Iterative Optimization Approach for Unified Image Segmentation and Matting". Jue Wang, Michael Cohen