出版社：National Defense University Barbaros Naval Sciences and Engineering Institute Journal of Naval Science and Engineering
摘要：In this study,we propose to detect interest regions based on high level image features.For this purpose,Maximally Stable Extreme Regions (MSER) approach is extended using the texture information of image pixels.After the image is convolved with the filters of various scales and orientation,each pixel is represented by the responses to those filters.The regions with different textures than their vicinity are detected using high dimensional pixel features.The proposed approach is evaluated in terms of repeatibility and matching scores using an experimental setup in the literature.It outperfoms the intensity and colour based studies,especially in the images containing textured regions.
其他摘要：Bu çalışmada,çok sayıda bilgisayarla görme uygulamasında kullanılan ilgi bölgelerinin üst düzey imge özniteliklerine dayalı olarak tespit edilmesi önerilmektedir.Bu amaçla,Maksimum Durağan Uç Bölgeler yaklaşımı,imge noktalarının desen bilgisi kullanılarak
