Bad expossition detection
When you return from a trip, you usually have a lot of repeated photos. There are lots of blurry photos, noise ones, and even some of them have awful exposition. Selecting the best ones is a very time consuming task, so we wanted to solve that issue.
What it does
It's a computer software that detects similar images from a input set and it selects the best ones automatically.
How I built it
The application is built with matlab and it uses different image processing technics such as filtering, frequency transformations, hierarchical clustering and colour feature extraction.
I build 4 algorithms to detect the independent features of a bad quality image such as blur, noise, bad exposition and desaligments. The main application detects similar images and discards the worse ones based on the features above.
Challenges I ran into
The design of every feature was a big challenge as it required some advanced signal processing concepts. The clustering application was also a bit tricky because it required some algorithms to process every cluster individually.
Accomplishments that I'm proud of
The performance of the individual algorithms, they run very well.
What I learned
I learned new different ways to retrieve important information from an image, and expanded my clustering algorithms knowledge in matlab.
What's next for ImaShare
A mobile app build in react-native is in the way, with the intention of using the algorithm with the purpose of facilitating the image sharing. The app would select automatically the best photos of an event and then share it with the selected people. The app will have integration with social platforms like Facebook.