confusion in VL_SLIC , outpur varies depending on how image is read

Hi, I am using VL_SLIC to implement segmentation. However the outputs vary depending on how I read the image.

Way1: Read the img as im2double , so the intensity ranges from 0 to 1. Here, the output superpixels are more regular and spacious.

%Input
img = im2double(imread('Images/2_14_s.bmp'));
regionSize = 30 ;
regularizer = 0.1;

%get the SLIC superpixels
vl_setup
segments1 = vl_slic(single(img), regionSize, regularizer) ;

Way2: Read the img as uint8 , so the intensity ranges from 0 to 255. Here, the output superpixels are more irregular but tightly fit to the edges.

%Input
img = imread('Images/2_14_s.bmp');
regionSize = 30 ;
regularizer = 0.1;

%get the SLIC superpixels
vl_setup
segments1 = vl_slic(single(img), regionSize, regularizer) ;

here are the images:

http://swagatikapanda-vision.blogspot.in/2012/10/confusion-in-vlslic.html

I want to know which is the right way? and why this change is affecting the performance?

[Walter Roberson]

This is not depending on how the image is read: this is depending on how you treat the data after it is read.

[Swagatika]

Can you please elaborate?

Answer 1

[Image Analyst]

After a search, it appears that vl_slic() is in a third party package available at http://www.vlfeat.org/overview/slic.html. It looks like regularizer needs to be changed by 255 to get the same values in the two methods because the input images differ by a factor of 255 in the two cases.

Answer 2

[Walter Roberson]

In both cases you are using imread(). That accomplishes all of the image reading. Using im2double() is processing after the image has been read.

MATLAB routines can generally handle images represented in several different ways, and the different ways are distinguished internally by examining the datatype (e.g., double or uint8) When the datatype is double then the data must be in the range 0 to 1, and the routines are permitted to malfunction if the data is some other range instead. When the datatype is uint8 then the data range is the integers 0 to 255. When you double() a uint8 matrix and pass the result into a MATLAB image routine, the image routine is going to get confused.

Answer 3

[Dark]

I have one question since this is concerning VL_SLIC. I am also using the VLFeat package. My question is how do I view the image with superpixels ? I tried imshow(segments) but just get a black image.

[Image Analyst]

It's probably a floating point array, so use []:

imshow(segments, []);

Answer 4

[ibtissem maidi]

when i use this code

segments = vl_slic(img,regionSize, regularizer)

i have this error :

Undefined function 'vl_slic' for input arguments of type 'double'."

i don't know where is a matter some one can help me please !

[Walter Roberson]

You need to install https://www.mathworks.com/matlabcentral/fileexchange/47809-vlfeat-vlfeat

[Image Analyst]

Or instead, simply just use the superpixels() function built in to the Image Processing Toolbox: https://www.mathworks.com/help/images/ref/superpixels.html?s_tid=srchtitle