Recognizing human actions in images using segment and contour features

In this paper, the aim is to recognize human actions from still images. Different from other works, the aim is to use a contour detection and an image segmentation algorithm together to see if the contours and segments are able to represent the human pose in a still image. As a solution, firstly, contours and segments are found in the image. The segments and contours are then used to extract the features. After extracting the features, the contoured and segmented images are split as a grid to get the spatial information. Then, the features in every grid are concatenated as a histogram and the histogram is used for classification. Support Vector Machines (SVM) algorithm is used for classification. The results demonstrate that using the contours and segments together in the image is a better representation for recognizing human actions in still images.