Çѱ¹Á¤º¸Åë½ÅÇÐȸ ³í¹®Áö (Journal of the Korea Institute of Information and Communication Engineering)
Current Result Document : 1 / 2
ÇѱÛÁ¦¸ñ(Korean Title) |
Sparse-Neighbor ¿µ»ó Ç¥Çö ÇнÀ¿¡ ÀÇÇÑ ÃÊÇØ»óµµ |
¿µ¹®Á¦¸ñ(English Title) |
Super Resolution by Learning Sparse-Neighbor Image Representation |
ÀúÀÚ(Author) |
¾ö°æ¹è
ÃÖ¿µÈñ
ÀÌÁ¾Âù
Kyoung-Bae Eum
Young-Hee Choi
Jong-Chan Lee
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 18 NO. 12 PP. 2946 ~ 2952 (2014. 12) |
Çѱ۳»¿ë (Korean Abstract) |
Ç¥º» ±â¹Ý ÃÊÇØ»óµµ(Super Resolution ÀÌÇÏ SR) ¹æ¹ýµé Áß ³×À̹ö ÀÓº£µù(Neighbor Embedding ÀÌÇÏ NE) ±â¹ýÀÇ ±âº» ¿ø¸®´Â Áö¿ªÀû ¼±Çü ÀÓº£µùÀ̶ó´Â ¸Å´ÏÆúµå ÇнÀ¹æ¹ýÀÇ °³³ä°ú °°´Ù. ±×·¯³ª, ³×À̹ö ÀÓº£µùÀº ±¹ºÎ ÇнÀ µ¥ÀÌÅÍ ÁýÇÕÀÇ Å©±â°¡ ³Ê¹« À۱⠶§¹®¿¡ ÀÌ¿¡ µû¸¥ ºó¾àÇÑ ÀϹÝÈ ´É·ÂÀ¸·Î ÀÎÇÏ¿© ¾Ë°í¸®ÁòÀÇ ¼º´ÉÀ» Å©°Ô ÀúÇϽÃŲ´Ù. º» ³í¹®¿¡¼´Â ÀÌ¿Í °°Àº ¹®Á¦Á¡À» ÇØ°áÇϱâ À§Çؼ ÀϹÝÈ ´É·ÂÀÌ ¶Ù¾î³ Support Vector Regression(ÀÌÇÏ SVR)À» ÀÌ¿ëÇÑ Sparse-Neighbor ¿µ»ó Ç¥Çö ÇнÀ ¹æ¹ý¿¡ ±â¹ÝÇÑ »õ·Î¿î ¾Ë°í¸®ÁòÀ» Á¦¾ÈÇÏ¿´´Ù. ÀúÇØ»óµµ ÀÔ·Â ¿µ»óÀÌ ÁÖ¾îÁö¸é bicubic º¸°£¹ýÀ» ÀÌ¿ëÇÏ¿© È®´ëµÈ ¿µ»óÀ» ¾ò°í, ÀÌ È®´ëµÈ ¿µ»óÀ¸·ÎºÎÅÍ ÆÐÄ¡¸¦ ¾òÀº ÈÄ ÀúÁÖÆÄ ÆÐÄ¡ÀÎÁö °íÁÖÆÄ ÆÐÄ¡ ÀÎÁö¸¦ ÆǺ°ÇÑ ÈÄ °¢ ¿µ»ó ÆÐÄ¡ÀÇ °¡ÁßÄ¡¸¦ ¾òÀº ÈÄ µÎ °³ÀÇ SVRÀ» ÈÆ·ÃÇÏ¿´À¸¸ç ÈÆ·ÃµÈ SVRÀ» ÀÌ¿ëÇÏ¿© °íÇØ»óµµÀÇ ÇØ´ç È¼Ò °ªÀ» ¿¹ÃøÇÏ¿´´Ù. ½ÇÇèÀ» ÅëÇÏ¿© Á¦¾ÈµÈ ±â¹ýÀÌ ±âÁ¸ÀÇ º¸°£¹ý ¹× ³×À̹ö ÀÓº£µù ±â¹ý µî¿¡ ºñÇØ Á¤·®ÀûÀΠôµµ ¹× ½Ã°¢ÀûÀ¸·Î Çâ»óµÈ °á°ú¸¦ º¸¿© ÁÖ¾ú´Ù. |
¿µ¹®³»¿ë (English Abstract) |
Among the Example based Super Resolution(SR) techniques, Neighbor embedding(NE) has been inspired by manifold learning method, particularly locally linear embedding. However, the poor generalization of NE decreases the performance of such algorithm. The sizes of local training sets are always too small to improve the performance of NE. We propose the Learning Sparse-Neighbor Image Representation baesd on SVR having an excellent generalization ability to solve this problem. Given a low resolution image, we first use bicubic interpolation to synthesize its high resolution version. We extract the patches from this synthesized image and determine whether each patch corresponds to regions with high or low spatial frequencies. After the weight of each patch is obtained by our method, we used to learn separate SVR models. Finally, we update the pixel values using the previously learned SVRs. Through experimental results, we quantitatively and qualitatively confirm the improved results of the proposed algorithm when comparing with conventional interpolation methods and NE. |
Å°¿öµå(Keyword) |
ÃÊÇØ»óµµ
Sparse-Neighbor ¿µ»ó Ç¥Çö
ÁöÁö º¤ÅÍ È¸±Í
Super Resolution
Sparse-Neighbor Image Representation
Support Vector Regression
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|