Shoeprint Database 
Database images were collected from the invited participants which is called Iranian-Turkish Shoeprints database (ITSP DB). For gathering the images, at first, a form was printed at the top of 1000 A3-size (297×420 mm) paper sheets which was aimed at recording participants’ information with regard to the shoes. The recorded information included features about height, weight and the gender of the participants, style, size and being left or right shoe. It should be mentioned that all the shoeprints were recorded from the participants’ right shoes. In recording and registering shoeprints, the participants firstly pressed their right foot on the inkpad so that the shoe sole is impregnated with ink. Next, the participants were asked to press their right foot on the intended paper; Furthermore, each participant’s general demographic information were recorded in the form at the top of the paper. However, it should be pointed out that nothing about the identity of the participant was recorded in the forms; that is, the forms were anonymous. A notable challenge and problem regarding the production of the database was that sometime many participants were wearing the same or highly similar shoes. Hence, for removing this problem, five separate samples were recorded from each shoe. Finally, shoeprint papers were scanned with the resolution of 300 dots per inch (DPI) and with a colored surface. Then, they were converted into the digital files. Since some shoeprint images were not practically useable, the recorded shoeprint images were divided into 190 classes with five images within each class. Hence, a total of 950 shoeprint images (190×5=950) were obtained. Then, the forms printed at the top of the paper sheets for storing each participant’s information were manually removed from the upper side of the paper sheets. 

All 190 classes of shoeprints had patterns which were different from one another. All 5 images within each class have been taken from the same shoe at different sessions. In the conducted simulations, only two samples of them were used. Out of 190 classes of shoeprint images, one image was registered in the reference database and one image was registered in the query database. Hence, the reference database included 190 shoeprint images and the query database included 190 images. In other words, 190 images were used as queries. In producing database, we tried to achieve a real state of the shoeprints which would be highly similar to the actual shoe marks taken from crime scenes. Since one dominant feature of actual shoeprints obtained from crime scenes is that they are not complete forms of the shoeprints in crime scenes, hence, we focused on utilized (second-hand) shoes in producing the database.