how to view a convolved image in matlab?

7 次查看（过去 30 天）
sunkara manoj
sunkara manoj 2017-6-5
回答： Image Analyst 2017-6-5
I am currently working on a project where I need to convolve two images grayscale images (0-255 range).I am using conv2() command but the output image matrix is having very high intensity value (292643 and above). I have tried normalizing and then convoluting but that did not work. Here is the code and the image I used.
clc;
clear all;
close all;
image=imread('du5.jpg');
figure(),imshow(image);
grayimage= rgb2gray(image);
im=grayimage;
k=imerode(im,strel('disk',1));
x=imdilate(k,strel('disk',1));
figure(),imshow(x);
k = 1000;
st=std2(x);
sigma1 = st/10;
sigma2 =sigma1*k;
h1 = fspecial('gaussian', [3,3], sigma1);
h2 = fspecial('gaussian', [9,9], sigma2);
gauss1 = imfilter(x,h1,'replicate');
gauss2 = imfilter(x,h2,'replicate');
dogImg = gauss1 - gauss2;
i=conv2(dogImg,im);

请先登录，再回答此问题。

采纳的回答

Image Analyst
Image Analyst 2017-6-5
The whole algorithm makes no sense. So you're getting the spatial spread of the Gaussian from the standard deviation of the intensity??? One is a distance and the other is a gray level. That is nonsense. They're different units! I see no justification for using an intensity to set a spatial parameter for finding edges.
Next, you convolve the Gaussian with the image using imfilter(). Okay, that's fine. And then you subtract them to get the dog filtered image. Okay, again that's fine. If you just wanted an edge detection image, you should be done now. But to then convolve the filtered image with the original??? What's the logic behind that? Even if you did, you'd end up with an output image 4 times the size of your original images. What are you trying to achieve?
You start off by doing a morphological opening (erosion followed by dilation, which could be done in one step by imopen()). Why? This will essentially expand small dark blobs. Why do you want to do that?
So for multiple reasons I think you need to rethink what you're trying to do.

更多回答（1 个）

Walter Roberson
Walter Roberson 2017-6-5
https://www.mathworks.com/help/matlab/ref/conv2.html#inputarg_shape
Shape:
Subsection of the convolution, specified as one of these values:
  • 'full' — Return the full 2-D convolution.
  • 'same' — Return the central part of the convolution, which is the same size as A.
  • 'valid' — Return only parts of the convolution that are computed without zero-padded edges.
  1 个评论
Walter Roberson
Walter Roberson 2017-6-5
dogImg = im2double(gauss1) - im2double(gauss2);
i = conv2(dogImg, im2double(im)); %-163 to +141
figure()
imshow(i)
figure()
i8 = uint8(i);
imshow(i8);

请先登录，再进行评论。

类别

在 Help Center 和 File Exchange 中查找有关 Object Analysis 的更多信息

标签

产品

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by