榫欐簮鏈熷垔缃?/p>
http://www.qikan.com.cn
涓�涓熀浜庤秴鍍忕礌鐨勫浘鍍忓垎鍓茬畻娉?/p>
浣滆�咃細鍒樹箻椋?/p>
寰愬垰
鏉ユ簮锛氥�婄數鑴戠煡璇嗕笌鎶�鏈�?/p>
2017
骞寸
25
鏈?/p>
鎽樿锛氬浘鍍忓垎鍓叉妧鏈槸璁＄畻鏈鸿瑙夐鍩熺殑鍩虹宸ヤ綔锛岀洰鍓嶇殑鍥惧儚鍒嗗壊鎶�鏈ぇ澶氫互鍍忕礌涓哄崟
浣嶈繘琛屻�傞殢鐫�鍥惧儚鍒嗚鲸鐜囩殑鎻愬崌锛屽浘鍍忓垎鍓茬殑鎴愭湰瓒婃潵瓒婇珮锛屼簬鏄鎵鹃檷浣庡浘鍍忓垎鍓叉垚鏈殑闇�
姹傝秺鏉ヨ秺杩垏銆傚熀浜庤秴鍍忕礌鐨勫浘鍍忓垎鍓茬畻娉曞氨鏄繖鏍蜂竴绉嶆柟娉曘�傝秴鍍忕礌鍏跺疄灏辨槸灏嗗叿鏈夌浉浼肩汗
鐞嗐�侀鑹层�佷寒搴︾瓑鐗瑰緛鐨勭浉閭诲儚绱犺繘琛屽悎骞跺悗鐨勫浘鍍忓潡銆傞噰鐢ㄨ秴鍍忕礌鐨勫浘鍍忓垎鍓茬畻娉曞悗锛屽彲浠?/p>
娑堥櫎鍍忕礌闂寸殑鍐椾綑锛屽ぇ澶ч檷浣庡悗缁浘鍍忓鐞嗕换鍔＄殑鎴愭湰锛屽姞蹇浘鍍忚瘑鍒殑閫熷害銆?/p>
鍏抽敭璇嶏細璁＄畻鏈鸿瑙夛紱
鍥惧儚鍒嗗壊锛?/p>
瓒呭儚绱狅紱
K-means
鑱氱被锛?/p>
褰掍竴鍖栧垎鍓?/p>
涓浘鍒嗙被鍙凤細
TP317
鏂囩尞鏍囪瘑鐮侊細
A
鏂囩珷缂栧彿锛?/p>
1009-3044
锛?/p>
2017
锛?/p>
25-0165-02
Abstract
锛?/p>
Image segmenting is the basic technology in computer vision
锛?/p>
and most image
segmenting technology is Based on pixels in nowadays. With the promoting of image resolution
锛?/p>
the
cost of image segmenting is higher and higher
锛?/p>
so the requirement of lower image segmenting
method is more and more urgent. Image segmenting Based on superpixel is such method. In fact
superpixel is an image block consisting of neighbor pixels
锛?/p>
which has the similar texture
锛?/p>
color
锛?/p>
and brightness. By taking superpixel method
锛?/p>
it will eliminate redundancy of pixels
锛?/p>
and decrease
the cost of following image processing
锛?/p>
and accelerate image recognizing.
Key words
锛?/p>
computer vision
锛?/p>
image segmenting
锛?/p>
superpixel
锛?/p>
K-means clustering
锛?/p>
normalized cut
1
姒傝堪
鍥惧儚鐗瑰緛鍙嶆槧浜嗗浘鍍忕殑鍩烘湰鐗规�э紝涓�鑸彲鍒嗕负鍏ㄥ眬鐗瑰緛銆佸眬閮ㄧ壒寰佸拰绌洪棿鍏崇郴鐗瑰緛銆?/p>
鍏ㄥ眬鐗瑰緛涓昏鏈夐鑹茬壒寰佸拰绾圭悊鐗瑰緛銆傞鑹茬壒寰佸拰绾圭悊鐗瑰緛鎻忚堪浜嗗浘鍍忔墍瀵瑰簲鐨勫璞＄殑琛?/p>
闈㈡�ц川銆?/p>
灞�閮ㄧ壒寰佷富瑕佹槸杞粨鐗瑰緛鍜屽尯鍩熺壒寰併�傝疆寤撶壒寰佷富瑕佹弿杩颁簡鐗╀綋鐨勫杈圭晫锛屽尯鍩熺壒寰佸垯鎻?/p>
杩版暣涓舰鐘跺尯鍩熴�?/p>
绌洪棿鍏崇郴鐗瑰緛鏄寚鍥惧儚涓垎鍓插嚭鏉ョ殑澶氫釜鐩爣涔嬮棿鐨勭浉浜掔殑绌洪棿浣嶇疆鎴栫浉瀵规柟鍚戝叧绯汇�傜┖
闂村叧绯诲張鍒嗕负涓ょ被锛氱浉瀵圭┖闂翠綅缃俊鎭拰缁濆绌洪棿浣嶇疆淇℃伅銆傜浉瀵圭┖闂翠綅缃叧绯诲己璋冪殑鏄洰鏍?/p>
涔嬮棿鐨勭浉瀵规儏鍐碉紝濡備笂涓嬪乏鍙崇瓑锛岀粷瀵圭┖闂翠綅缃俊鎭叧绯诲己璋冪殑鏄洰鏍囦箣闂寸殑璺濈澶у皬浠ュ強鏂?/p>
浣嶏紝濡傞偦鎺ャ�侀噸鍙犮�佸拰鍖呭绛夈�?/p>