Ȩ > Ç¥ÁØÈ °³¿ä > TTAÀÇ Ç¥ÁØÇöȲ
Ç¥ÁعøÈ£ | TTAK.KO-10.1387 | ±¸Ç¥ÁعøÈ£ | |
---|---|---|---|
Á¦°³Á¤ÀÏ | 2022-12-07 | ÃÑÆäÀÌÁö | 20 |
ÇѱÛÇ¥Áظí | ¶óÀÌÆ® Çʵå(Light Field) À̹ÌÁöÀÇ ¿ÃÀÎÆ÷Ä¿½º(All-in-Focus)¸¦ À§ÇÑ ÇнÀ ±â¹Ý ½Å°æ¸Á ±¸Á¶¿Í ¼º´É Æò°¡ ¹æ¹ý | ||
¿µ¹®Ç¥Áظí | Learning Based Neural Network Structure and Performance Evaluation Method for All-in-Focus of Light Field Images | ||
Çѱ۳»¿ë¿ä¾à | ÀÌ Ç¥ÁØÀº ÀԷ´ÜÀ¸·Î ¶óÀÌÆ® Çʵå À̹ÌÁö¸¦ ÀÌ¿ëÇÏ°í ½Å°æ¸ÁÀº Ư¡ ÃßÃâ, Ư¡ ÇÕ¼º, Ư¡ À籸¼º ¸ðµâ·Î ³ª´©¾îÁ® ÀÖÀ¸¸ç Ư¡ ÃßÃâ ¸ðµâÀº ´ÙÁß ÇØ»óµµ º´·Ä ½Å°æ¸Á ±¸Á¶¸¦ ÀÀ¿ëÇØ À̹ÌÁö¿¡ ´ëÇÑ Æ¯Â¡ ÃßÃâÀ» ÇÏ°Ô µÈ´Ù. ÀÌ·¸°Ô ÃßÃâµÈ Ư¡¸ÊÀº Ư¡ ÇÕ¼º ¸ðµâ, Ư¡ À籸¼º ¸ðµâÀ» Åë°úÇϸç Á¾´Ü°£ ÇнÀ(End-to-End)¹æ½ÄÀ¸·Î ¿ÃÀÎÆ÷Ä¿½º À̹ÌÁö¸¦ ¸¸µé¾î³½´Ù. ¿ÃÀÎÆ÷Ä¿½º À̹ÌÁö¿¡ ´ëÇÑ ¼º´É Æò°¡ ¹æ¹ý ¶ÇÇÑ Á¤ÀÇÇÑ´Ù. | ||
¿µ¹®³»¿ë¿ä¾à | The standard uses refocused image of Light Field as input, and the neural network is divided into feature extraction, feature fusion, and feature reconstruction modules, which apply multi scale resolution parallel neural network structures to extract features for refocused images. The feature map extracted in this way passes through the feature fusion module and feature reconstruction module and creates All-in-Focus images in the End-to-End method. We also define performance evaluation methods for All-in-Focus images. | ||
±¹Á¦Ç¥ÁØ | |||
°ü·ÃÆÄÀÏ | TTAK.KO-10.1387.pdf |