使用Julia进行图像处理--图像表示与匹配算法

img_rosenfield[dilate(imcorner(img_f, method=kitchen_rosenfeld)) .> 0.01] .= colorant"yellow"

img_fast = copy(img)

img_fast[dilate(fastcorners(img_f, 12, 0.05)) .> 0.01] .= colorant"yellow"

通常，imcorner和fastcorners的结果很小，例如一个像素。从ImageMorphology软件包中应用了膨胀函数，以突出显示通过imcorner或FAST确定的结果，可以让用户在绘制和预览特征时更好地表示特征。还使用了着色剂技术来获得精确的颜色值。

将所有四个图像合并为一个图像：

new_img = vcat(

hcat(img_harris, img_shi),

hcat(img_rosenfield, img_fast)

)

new_img[Int(size(new_img, 1) / 2), :] .= colorant"yellow"

new_img[:, Int(size(new_img, 2) / 2)] .= colorant"yellow"

imshow(new_img)

尝试在猫的图像上运行相同的代码，然后再次比较结果以查看是否有任何变化：

using Images, ImageFeatures, ImageMorphology, ImageView

img = Gray.(restrict(load(“cat-3417184_640.jpg”)))

img_f = Float16.(Gray.(img))

img_harris = copy(img)

img_harris[dilate(imcorner(img_f, method=harris)) .> 0.01] .= colorant"yellow"

img_shi = copy(img)

img_shi[dilate(imcorner(img_f, method=shi_tomasi)) .> 0.01] .= colorant"yellow"

img_rosenfield = copy(img)

img_rosenfield[dilate(imcorner(img_f, method=kitchen_rosenfeld)) .> 0.01] .= colorant"yellow"

img_fast = copy(img)

img_fast[dilate(fastcorners(img_f, 12, 0.05)) .> 0.01] .= colorant"yellow"

new_img = vcat(

hcat(img_harris, img_shi),

hcat(img_rosenfield,img_fast)

)

imshow(new_img)

除FAST以外的所有方法在图像周围都有角点。原因是，imcorner函数返回所有角点而没有任何阈值参数。

更新阈值以仅检索特征的前5％，然后再次比较结果：

img = Gray.(restrict(load(“cat-3417184_640.jpg”)))

img_f = Float16.(Gray.(img))

img_harris = copy(img)

img_harris[dilate(imcorner(img_f, Percentile(95), method=harris)) .> 0.01] = colorant"yellow"

img_shi = copy(img)

img_shi[dilate(imcorner(img_f, Percentile(95), method=shi_tomasi)) .> 0.01] = colorant"yellow"

img_rosenfield = copy(img)

img_rosenfield[dilate(imcorner(img_f, Percentile(95), method=kitchen_rosenfeld)) .> 0.01] = colorant"yellow"

img_fast = copy(img)

img_fast[dilate(fastcorners(img_f, 12, 0.05)) .> 0.01] = colorant"yellow"

new_img = vcat(

hcat(img_harris, img