Genetic algorithm in boosting for object class image segmentation

We describe how a task in computer vision can be effectively resolved by employing Genetic Algorithm. This paper focuses on the problem of semantic segmentation of digital images. We propose to use an improved genetic algorithm for the learning parameters of weak classifiers in a boosting learning set up. We propose a new encoding and genetic operators in accordance with this problem. Beside that, we employed multiple image features such as texture-layout, location, color and HoG for improving the accuracy of the system. Experiments are conducted extensively on MSRC, a widely used benchmark image datasets. The experimental results demonstrate that the performance of our system is comparable to, or even outperforms the state-of-the-art algorithms in semantic segmentation.