Consider a 3 × 3 spatial mask that averages the intensity of all neighbours of a pixel (x, y) in this 3 × 3 neighbourhood, but excluding the point itself.

4 次查看（过去 30 天）

显示更早的评论

Ben Ma 2021-9-8

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1449094-consider-a-3-x-3-spatial-mask-that-averages-the-intensity-of-all-neighbours-of-a-pixel-x-y-in-thi

评论： Rena Berman 2021-9-23

采纳的回答： Image Analyst

How to make a function that applies this filter to an image and displays the original image and the filtered image?

Thanks.

2 个评论
显示无隐藏无

Image Analyst 2021-9-9

Ben, stop destroying your questions or people will stop helping you.

Rena Berman 2021-9-23

(Answers Dev) Restored edit

请先登录，再进行评论。

请先登录，再回答此问题。

采纳的回答

Image Analyst 2021-9-8

3
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1449094-consider-a-3-x-3-spatial-mask-that-averages-the-intensity-of-all-neighbours-of-a-pixel-x-y-in-thi#answer_783374

在 MATLAB Online 中打开

Make a kernel that has all 1's except for the center, then divide by 8:

kernel = ones(3,3) / 8;
kernel(2, 2) = 0;
blurredImage = conv2(double(grayImage), kernel, 'same');

There will be some slight errors in the very top, left, top, and bottom row. If you need to correct those you can use two calls to conv2() - one to count pixels and one to sum up values, then divide those two images pixel-by-pixel.

3 个评论
显示 1更早的评论隐藏 1更早的评论

Image Analyst 2021-9-8

Yes, though imfilter will round to the nearest gray level if it's a uint8 image. You won't get fractional numbers.

Image Analyst 2021-9-9

Ben:

For your original question where you wanted the pixel to be replaced by the average of the 8 surrounding neighbors, you'd divide by 8.

For your edited question, which does not specify which pixels to include in the average (Why did you change your question???) you should divide by the number of pixels in the averaging window. If it's a 3x3 window and includes the central pixel, that's 9 pixels so you should divide the sum by 9. For a 5x5 using all pixels, divide by 5x5 = 25.

请先登录，再进行评论。