Abstract: Seam-carving is one of the widely used content aware image resizing methods. According to the image content, the output of the algorithm sometimes may not be satisfactory visually. In the presented study a machine learning method for detecting visually impaired images after seam-carving procedure is proposed. For this purpose a training set that is constructed by good and visually bad examples which are obtained by examining the seam-carving images. Features from the sample set are extracted and a Support Vector Machine is trained for determining whether the quality of the examined image is good. According to experimental results using four fold cross-validation, the method produced 65% success.

Keywords: Seam-carving, Support Vector Machine, Image Resizing, Local Binary Patterns.