Detecting facial features automatically

Abstract

Obje tespit etmek icin bir cok algoritma ve yaklasm vardr. Bunlarn cogu Viola Jonesalgoritmasna dayanr. Bizim edindigimiz tecrubelere gore, obje tespitinde temelkonu o objeye ait ozniteliklerdir. Bu oznitelikler genellikle manuel olarak olusturulur.Bu tezde biz Haar-like ozniteliklerin otomatik ckarmlar uzerine arastrma yaptk.Gri tonlamal resimler icin tamamyla otomatiklestirilmis bir yuz alglaycs tasarlayp bunu uyguladk. Nihayetinde, tasarladgmz algoritmann farkl veribankalaruzerindeki performansn gosterdik.

There are many algorithms and approaches in object detection world. Many of themare based on Viola Jones algorithm. According to our observations, the features whichhelp to detect an object are very critical for the success of this algorithm. These featuresare usually created manually. In this thesis we explore automatic extractionof Haar-like features. We describe the design and construction of a completely automatedface detector for gray scale images. Finally, we illustrate the performance ofour algorithm on various databases.