Fragment image into MxN regions and calculate a texture weight for each one and then find corners

1 次查看（过去 30 天）

Katerina 2014-8-19

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/151503-fragment-image-into-mxn-regions-and-calculate-a-texture-weight-for-each-one-and-then-find-corners

评论： Katerina 2014-8-19

What i have to do really is an implementation of harris corner method..but not the traditional one.I tried to write the steps clear (well,thats what i understand):

-Fragment the original image(1280x720) into MxN regions. e.g 80x80

-Do some stuff to calculate the normalized variance of each region. (output size of array: regions x 1)?

-Then to determine the weight of that region, calculate w(i) (output same size)

- w(i) is used to determine the number of corners B(i) for each region. where Sum is the predefined total number of corners for all regions e.g 6. B(i)=w(i) x Sum (what does this mean??)

-R value is calculated for each pixel in each region (R = det(M) - k * (trace(M) ^ 2); from harris corner method).I used Peter Kovesi implementation of harris. (diferrent size of matrix R for each region)?

- All pixels in each region are sorted in a descending order of their R values. The first Bi pixels in each region are then selected as corners .How B(i)=w(i) x Sum can be selected as corners? Will i have to do something like compare with R values?

Sο..except the above,my main questions are:

How can i fragment the image and be able to handle each region, and finally use all R values of all regions to figure the corners detected? I tried blockproc..divide,split,cells methods,and other methods and demos but i think i missunderstood something.How from each iteration of a blocked image 80x80 that computes R values(therefore pixel coordinates) i will put all these to original image that is 1280x720? Hope someone understand me a little i loose my brain cells..

2 个评论
显示无隐藏无

Image Analyst 2014-8-19

Exactly what does "be able to handle each region" mean? blockproc() gives you a sub-rectangle of the image. What you do with it (handle it) is up to you.

Katerina 2014-8-19

so you mean, blockprock will give me regions of size MxN-sub-rectangles of the image-

What will i use as the input function that blockprock requires?

i want for each block to find the corners through the procedure described and 'compose' all corners from all blocks to the original image in a figure.

请先登录，再进行评论。

请先登录，再回答此问题。