ImagePyramid
ImagePyramidApply
GradientFilter
ImageSaliencyFilter
SetAlphaChannel

Image Processing
Computer Vision

Multiscale Feature Detection

Apply gradient filtering or saliency filtering to images

Open in Cloud
Download Notebook

Start with an image:

In[1]:=

img=

;

Apply a gradient filter at several scales simultaneously:

In[2]:=

pyramid=ImagePyramid[img,"Gaussian",4]

Out[2]=

ImagePyramid

Type: Gaussian Levels: 4
Dimensions: {480,320} ↔ {60,40}



Apply gradient filtering to all levels of the pyramid:

In[3]:=

pyramidG=ImagePyramidApply[imgGradientFilter[img,1.4],pyramid];

Reconstruct assuming a Laplacian pyramid to add up features extracted from all levels:

In[4]:=

InverseImagePyramid[pyramidG,"Laplacian"]//ImageAdjust

Out[4]=

Compare the result with the gradient filter at a single fixed scale:

In[5]:=

GradientFilter[img,1.4]//ImageAdjust

Out[5]=

Next, consider a different image to compute interesting or important regions of the image using multiscale saliency filtering:

In[6]:=

img=

;

Apply a saliency filter to all levels of an image pyramid to get a multiscale result:

In[7]:=

pd=ImagePyramid[img,"Gaussian",4];

Apply saliency filtering to all levels of the pyramid:

In[8]:=

pdS=ImagePyramidApply[ImageSaliencyFilter,pd];

Reconstruct the image assuming a Laplacian pyramid to add up features extracted from all levels:

In[9]:=

SetAlphaChannel[img,InverseImagePyramid[pdS,"Laplacian"]//ImageAdjust]

Out[9]=

Compare with image saliency filtering at a single scale:

In[10]:=

SetAlphaChannel[img,ImageSaliencyFilter[img]]

Out[10]=

Related Symbols

ImagePyramid
ImagePyramidApply
GradientFilter
ImageSaliencyFilter
SetAlphaChannel

Publisher Information

Contributed by: Wolfram Staff

Wolfram Language Example Repository

Multiscale Feature Detection

See Also

Publisher Information