OpenCV图像插值方法的比较

最新推荐文章于 2024-05-23 19:30:22 发布

SimonCoder

最新推荐文章于 2024-05-23 19:30:22 发布

阅读量5.4k

点赞数 3

分类专栏： -----计算机视觉计算机视觉

本文链接：https://blog.csdn.net/simoncoder/article/details/89006673

版权

-----计算机视觉同时被 2 个专栏收录

7 篇文章 2 订阅

订阅专栏

计算机视觉

6 篇文章 0 订阅

订阅专栏

计算机视觉任务中，经常要使用图像插值方法来改变图像的尺寸，如图像金字塔、图像超分辨的预处理等，可以说图像插值方法是计算机视觉任务的基本操作。本文对opencv里的图像插值方法进行分析比较。

opencv里改变图像尺寸的函数是resize，下边是opencv2.3.14的resize函数的参考文档：

C++: void resize(InputArray src, OutputArray dst, Size dsize, double fx=0, double fy=0, int interpolation=INTER_LINEAR)

Parameters:
src – input image.
dst – output image; it has the size dsize (when it is non-zero) or the size computed from src.size(), fx, and fy; the type of dst is the same as of src.
dsize –
output image size; if it equals zero, it is computed as:

$\texttt{dsize = Size(round(fx*src.cols), round(fy*src.rows))}$

Either dsize or both fx and fy must be non-zero.

fx –
scale factor along the horizontal axis; when it equals 0, it is computed as

$\texttt{(double)dsize.width/src.cols}$

fy –
scale factor along the vertical axis; when it equals 0, it is computed as

$\texttt{(double)dsize.height/src.rows}$

interpolation –
interpolation method:

INTER_NEAREST - a nearest-neighbor interpolation
INTER_LINEAR - a bilinear interpolation (used by default)
INTER_AREA - resampling using pixel area relation. It may be a preferred method for image decimation, as it gives moire’-free results. But when the image is zoomed, it is similar to the INTER_NEAREST method.
INTER_CUBIC - a bicubic interpolation over 4x4 pixel neighborhood
INTER_LANCZOS4 - a Lanczos interpolation over 8x8 pixel neighborhood

Parameters:	src – input image. dst – output image; it has the size dsize (when it is non-zero) or the size computed from src.size(), fx, and fy; the type of dst is the same as of src. dsize – output image size; if it equals zero, it is computed as: $\texttt{dsize = Size(round(fxsrc.cols), round(fysrc.rows))}$ Either dsize or both fx and fy must be non-zero. fx – scale factor along the horizontal axis; when it equals 0, it is computed as $\texttt{(double)dsize.width/src.cols}$ fy – scale factor along the vertical axis; when it equals 0, it is computed as $\texttt{(double)dsize.height/src.rows}$ interpolation – interpolation method: INTER_NEAREST - a nearest-neighbor interpolation INTER_LINEAR - a bilinear interpolation (used by default) INTER_AREA - resampling using pixel area relation. It may be a preferred method for image decimation, as it gives moire’-free results. But when the image is zoomed, it is similar to the INTER_NEAREST method. INTER_CUBIC - a bicubic interpolation over 4x4 pixel neighborhood INTER_LANCZOS4 - a Lanczos interpolation over 8x8 pixel neighborhood

参考文档详细内容请参考：https://docs.opencv.org/2.4.13/modules/imgproc/doc/geometric_transformations.html?highlight=resize#cv2.resize。简单说明一下，src是输入图像，dst是输出图像，dsize是输出图像的尺寸，由用户指定，fx和fy是缩放因子，dsize与fx、fy两组参数只能选择一组使用，interpolation是可选的插值方法。

本文主要对opencv不同的插值方法进行比较，也就是INTER_NEAREST、INTER_LINEAR、INTER_AREA、INTER_CUBIC和INTER_LANCZOS4。实验所使用的图像为Art：

原始图像尺寸为1390×1110，选用的缩放因子为3，使用PSNR与SSIM作为评价标准。本文内容主要分为以下几个部分：

插值方法之间的相似度比较
插值方法下采样上采样的比较——BGR三通道
插值方法下采样上采样的比较——亮度通道Y，由CV_BGR2YUV转换
插值方法下采样上采样的比较——亮度通道Y，由CV_BGR2YCrCb转换

1、插值方法之间的相似度比较

对于原始尺寸1390×1110的Art图像，按照缩放因子3对原始图像进行下采样，得到低分辨率图像。使用不同的插值方法进行这一操作，比较低分辨率图像之间的相似程度，结果如下：

	nearest	bilinear	bicubic	area	lanczos4
nearest		27.261/0.850	27.002/0.843	28.300/0.882	26.994/0.842
bilinear			50.176/0.999	38.294/0.980	49.538/0.998
bicubic				37.007/0.975	55.615/0.999
area					36.966/0.975
lanczos4

由上表可知，插值方法bilinear、bicubic和lanczos4最相似度最高。

2、插值方法下采样上采样的比较——BGR三通道

对原始图像，按照缩放因子3对原始图像下采样再上采样，opencv读入彩色图像为BGR三通道图像，上采样下采样操作分别对每一个单通道进行。下表的结果是使用对应插值方法得到的结果与原始图像（ground truth）的比较结果：

down\|up	nearest	bilinear	bicubic	area	lanczos4
nearest	25.439/0.756	28.052/0.840	27.643/0.833	26.863/0.800	27.529/0.827
bilinear	30.019/0.869	33.044/0.917	33.429/0.920	31.278/0.890	33.281/0.915
bicubic	29.837/0.866	33.009/0.917	33.185/0.918	31.101/0.888	32.971/0.911
area	30.734/0.879	32.848/0.910	34.031/0.923	31.872/0.895	34.208/0.925
lanczos4	29.807/0.865	32.940/0.916	33.122/0.917	31.070/0.887	32.915/0.911

3、插值方法下采样上采样的比较——亮度通道Y，由CV_BGR2YUV转换

对原始图像，首先将读入的BGR彩色图像使用opencv的函数CV_BGR2YUV转换到YUV三通道，然后再将Y通道分离出来，仅对Y通道做下采样再上采样。下表的结果是对Y通道使用对应的插值方法得到的结果与原始图像的Y通道比较的结果：

down\|up	nearest	bilinear	bicubic	area	lanczos4
nearest	25.604/0.766	28.243/0.847	27.828/0.840	27.061/0.808	27.710/0.833
bilinear	30.192/0.874	33.161/0.920	33.517/0.923	31.464/0.894	33.365/0.917
bicubic	30.006/0.871	33.124/0.920	33.271/0.920	31.284/0.892	33.051/0.914
area	30.914/0.884	32.991/0.913	34.136/0.926	32.067/0.899	34.305/0.928
lanczos4	29.975/0.871	33.053/0.919	33.206/0.920	31.250/0.891	32.995/0.913

4、插值方法下采样上采样的比较——亮度通道Y，由CV_BGR2YCrCb转换

对原始图像，首先将读入的BGR彩色图像使用opencv的函数CV_BGR2YCrCb转换到YCrCb三通道，然后再将Y通道分离出来，仅对Y通道做下采样再上采样。下表的结果是对Y通道使用对应的插值方法得到的结果与原始图像的Y通道比较的结果：

down\|up	nearest	bilinear	bicubic	area	lanczos4
nearest	25.726/0.765	28.347/0.847	27.931/0.840	27.155/0.808	27.812/0.833
bilinear	30.271/0.874	33.213/0.920	33.565/0.923	31.514/0.894	33.409/0.918
bicubic	30.082/0.871	33.174/0.920	33.317/0.921	31.330/0.892	33.093/0.914
area	31.002/0.884	33.039/0.914	34.184/0.926	32.121/0.899	34.350/0.928
lanczos4	30.051/0.871	33.104/0.920	33.252/0.921	31.296/0.891	33.035/0.914

SimonCoder

关注

3
点赞
踩
10

收藏

觉得还不错? 一键收藏
0
评论
OpenCV图像插值方法的比较

计算机视觉任务中，经常要使用图像插值方法来改变图像的尺寸，如图像金字塔、图像超分辨的预处理等，可以说图像插值方法是计算机视觉任务的基本操作。本文对opencv里的图像插值方法进行分析比较。opencv里改变图像尺寸的函数是resize，下边是opencv2.3.14的resize函数的参考文档：C++:voidresize(InputArraysrc, OutputArrayds...
复制链接

扫一扫

专栏目录