HOG of Region of Interest for Improving Clothing Retrieval Performance

Due to spread of garment online store and the revenue from the online clothing market, clothing retrieval is becoming an active research topic in the last decade. The buyer commonly searches for shape, but the availability of various clothing designs and details on the Internet caused a growing interest to buy clothing with similar detail, at a particular position. The proposed method aims at enhancing the performance of garment image retrieval, in which the garments have similar details at a particular position. The method is developed based on histogram of oriented gradient (HOG), with a new approach for ROI partition mechanism prior employing the HOG. Simulations with four categories of clothing details confirmed that the proposed approach can improve the retrieval performance.