数字图像分析及应用.ppt

网络多媒体小组,数字图像分析及应用实例,2019/12/21,网络多媒体小组,2,内容提要,2019/12/21,2,数字图像概念,数字图像分析原理步骤,数字图像分析应用实例,2019/12/21,网络多媒体小组,3,数字图像概念,2019/12/21,3,数字图像：将模拟图象通过采样离散化得到数字图像，便于计算机处理。,2019/12/21,网络多媒体小组,4,介绍常见词,像素RGB灰色图像彩色图像,2019/12/21,网络多媒体小组,5,像素,像素：组成数字图像的元素。每一个像素的坐标对应特定的图像网格位置。像素的数值反映图像中对应位置的亮度。,像素,2019/12/21,网络多媒体小组,6,RGB,颜色空间之一-RGB,2019/12/21,网络多媒体小组,7,灰度图像（单色图像）,每个像素的亮度用一个数值来表示，通常数值范围在0到255之间（一个字节），0表示黑、255表示白，而其它表示灰度。,125,153,158,157,127,0,103,120,129,144,144,150,150,147,150,160,165,160,164,16,67,175,175,166,133,133,154,158,100,116,120,97,74,54,4,118,146,148,150,145,157,164,157,158,162,165,171,155,115,88,155,163,95,112,123,101,137,108,81,71,63,137,142,146,152,159,161,159,154,138,81,78,84,114,167,69,85,59,65,43,85,34,69,78,104,101,117,132,134,149,160,165,158,143,114,99,57,45,51,57,54,46,38,44,38,36,44,36,25,48,115,113,114,124,135,152,168,169,156,75,43,39,41,38,42,41,58,30,44,35,28,69,144,147,57,60,93,106,119,124,131,144,175,177,63,30,30,124,139,65,31,39,61,35,75,51,66,58,167,177,153,58,80,95,108,120,133,155,182,137,84,74,54,61,149,96,36,29,115,71,84,105,6,136,172,175,128,126,114,106,109,115,127,169,187,105,102,5,117,120,131,148,144,113,112,109,115,128,162,193,143,132,124,110,131,97,53,66,129,123,128,114,108,115,102,121,129,125,113,112,111,111,127,162,192,150,129,117,112,107,64,67,73,131,125,112,105,95,88,33,139,132,127,128,134,136,144,140,141,113,120,114,115,122,154,188,159,143,6,138,143,145,138,143,143,141,144,129,118,119,115,112,120,152,186,165,150,139,128,117,109,104,93,140,146,147,145,150,154,147,149,143,132,127,124,117,112,117,142,184,171,144,139,129,119,114,112,100135,143,145,151,150,143,146,142,139,131,131,127,119,115,114,140,190,177,141,135,133,121,120,115,102128,137,141,149,150,148,143,141,136,128,120,122,116,109,116,136,182,182,140,135,130,124,116,109,100131,132,141,143,148,147,143,140,134,125,116,117,108,109,120,130,171,185,133,132,127,120,117,110,36,131,139,136,140,141,140,136,132,121,114,117,109,106,112,130,163,192,130,126,131,132,132,136,140,134,135,128,117,110,113,111,110,112,129,151,196,155,130,127,120,115,103,64,127,132,130,128,134,134,140,132,126,110,102,117,118,114,112,121,152,195,165,123,128,117,115,96,36,123,131,127,129,132,133,134,131,127,112,109,117,114,116,107,116,145,172,1,2019/12/21,网络多媒体小组,8,彩色图像,(207,137,130)(220,179,163)(215,169,161)(210,179,172)(210,179,172)(207,154,146)(217,124,121)(226,144,133)(226,144,133)(224,137,124)(227,151,136)(227,151,136)(226,159,142)(227,151,136)(230,170,154)(231,178,163)(231,178,163)(231,178,163)(236,187,171)(236,187,171)(239,195,176)(239,195,176)(240,205,187)(239,195,176)(231,138,123)(217,124,121)(215,169,161)(216,179,170)(216,179,170)(207,137,120)(159,51,71)(189,89,101)(216,111,110)(217,124,121)(227,151,136)(227,151,136)(226,159,142)(226,159,142)(237,159,135)(237,159,135)(231,178,163)(236,187,171)(231,178,163)(236,187,171)(236,187,171)(236,187,171)(239,195,176)(239,195,176)(236,187,171)(227,133,118)(213,142,135)(216,179,170)(221,184,170)(190,89,89)(204,109,113)(204,115,118)(189,85,97)(159,60,78)(136,38,65)(160,56,75)(204,109,113)(227,151,136)(226,159,142)(237,159,135)(227,151,136)(230,170,154)(231,178,163)(230,170,154)(231,178,163)(231,178,163)(236,187,171)(239,195,176)(230,170,154)(204,109,113)(178,73,83)(216,179,170)(220,188,176)(190,77,84)(206,95,97)(217,113,113)(189,85,97)(226,144,133)(207,91,94)(170,56,72)(159,51,71)(150,47,67)(172,63,78)(226,144,133)(226,144,133)(227,151,136)(226,159,142),每个像素的亮度用一个一维向量来表示，各元素分别表示R、G、B通道，数值范围在0到255之间。,数字图像分析原理及步骤,2019/12/21,网络多媒体小组,9,2019/12/21,网络多媒体小组,10,预处理,消除模糊图像去噪图像增强水平矫正,图像分割的方法,1.基于边缘的分割方法先提取区域边界，再确定边界限定的区域。2.区域分割确定每个像素的归属区域，从而形成一个区域图。3.区域生长将属性接近的连通像素聚集成区域,2019/12/21,网络多媒体小组,11,基于边缘的分割方法,边缘检测算子（如梯度算子）检测出边缘，再利用光栅跟踪或全向跟踪将边缘像素连成线,2019/12/21,网络多媒体小组,12,区域分割,利用颜色灰度值判断某一像素点是否属于某特定区域,2019/12/21,网络多媒体小组,13,T（阈值）的确定取决于颜色分布直方图,2019/12/21,网络多媒体小组,14,区域分割,区域生长,选定某一像素点，并规定相似性判断标准，如果相邻像素点的相似度达到要求，就将区域有这一点向相邻点生长。直到所有相邻点都不符合相似性要求。,2019/12/21,网络多媒体小组,15,对象识别方法,模板匹配法神经网络法,2019/12/21,网络多媒体小组,16,模板匹配,2019/12/21,网络多媒体小组,17,输入图像,2019/12/21,网络多媒体小组,18,数字图像分析应用实例,军事公安方面指纹识别人脸鉴别交通监控生物医学工程方面CT技术癌细胞识别心电图分析航天和航空技术方面飞机遥感卫星遥感资源勘察,2019/12/21,18,人脸识别,2019/12/21,网络多媒体小组,19,模板匹配法人脸识别,2019/12/21,网络多媒体小组,20,待识别图像：图像库中图像：,相似性度量准则有：归一化互相关(NC)：平方差累加和(SSD)：绝对差累加和(SAD)：,识别结果,2019/12/21,网络多媒体小组,21,SAD差图像,2019/12/21,网络多媒体小组,22,Thankyou!,2019/12/21,网络多媒体小组,23,RGB颜色分布直方图,RGB颜色分布直方图：不同色彩在整幅图像中所占的比例，而并不关心每种色彩所处的空间位置。,2019/12/21,网络多媒体小组,24,梯度算子,用梯度算子对图像进行处理可以得到图像的边缘,