how to use image in exterma mfile?
显示 更早的评论
i download an mfile named exterma from here,which finds exterma points in an image, http://www.mathworks.com/matlabcentral/fileexchange/12275-extrema-m--extrema2-m/content/extrema/extrema.m and http://www.mathworks.com/matlabcentral/fileexchange/12275-extrema-m--extrema2-m/content/extrema/extrema2.m when I put my image address instead of xy,it gives me the error:(Undefined function 'extrema' for input arguments of type 'uint8'.) I don't know how to locate my image in this mfile,i just write these 2 lines: m=imread('image adress.jpg'); xy=rgb2gray(m); I also find 2 other mfiles here which detect exrema points in an image,but they only give me hundreds of 0 and 1,i want to detect exterma point in my image,show an image for result. http://www.mathworks.com/matlabcentral/fileexchange/45338-findextrema/content/findExtrema.m
http://www.mathworks.com/matlabcentral/fileexchange/41955-find-image-extrema isn't it better to use hit or miss for detecting exterma points?i want to detect top left,right top.... and for each one change the proper strel.?
采纳的回答
Iain
2014-9-18
Ok, I'll translate the problems:
Undefined function 'extrema' for input arguments of type 'uint8'.
That means that the extrema file is not visible to the matlab path. The command "which extrema.m" should result in a path being printed if it is on the path.
In terms of finding extrema on an image, you need to be very clear with what you mean.
Are you looking for the list of pixels which are higher than all of their neighbours, on a pixel to pixel basis, or are you looking for an area of bright (or dark) pixels (hotspots)?
If you need hotspots, this is something like what you want to do... If you take your image "im" and do this:
im = im - min(im(:));
im = im / max(im(:));
You'll have your image scaled between 0 and 1. You can then find hot spots/regions with this (modify the > 0.9 to suit)
hotspots = im > 0.9;
imagesc(hotspots)
figure
imagesc(hotspots .* im)
7 个评论
sara
2014-9-18
编辑：Image Analyst
2014-9-18
Thank you for your answer. I wanted to find corners like this image:
Iain
2014-9-18
Ok, so I've given you the first half then, and this is the second half:
labelled = bwlabel(hotspots);
stats = regionprops(labelled,'extrema');
sara
2014-9-20
Image Analyst
2014-9-20
Nope, won't work. Bytes have to multiply by bytes (uint8) not any other class. Plus dividing by the max may give you all zeros if it stays integer. Try this:
grayImage = imread('..');
m=rgb2gray(grayImage);
m = m - min(m(:));
hotspots = m > 0.9;
imagesc(hotspots)
figure
% Method 1: convert to double
m = double(m) / max(m(:));
maskedImage = hotspots .* m; % No need to cast because m is double now.
% Method 2, leave as integer, but will be all zeros or 1's.
maskedImage = m / max(m(:)); % Just 0 or 1s
maskedImage(~hotspots) = 0;
% Method 3, leave as integer, but will be all zeros or 1's.
maskedImage = m / max(m(:)); % Just 0 or 1s
maskedImage = maskedImage .* uint8(hotspots); % Cast logical to uint8
imshow(maskedImage, []);
labelled = bwlabel(hotspots);
stats = regionprops(labelled,'extrema');
sara
2014-9-20
Image Analyst
2014-9-20
You can use bwboundaries() on the mask image and then use a for loop to find the most distant points. See attached example. (Wow, I'm just posting tons of demos today. But only a fraction of all of them I have, which is about 150 of them.)
sara
2014-9-20
更多回答（0 个）
类别
在 帮助中心 和 File Exchange 中查找有关 ROI-Based Processing 的更多信息
Translated by 
