用2D索引矩阵索引4D数组 [英] Indexing a 4D array with a 2D matrix of indicies

问题描述

我目前有一个4D图像矩阵，其形式为高度x宽度x RGB x imageNumber，我想使用2D数组索引而不使用for循环. 2D数组的格式为高度x宽度，其值是要索引的图像编号.

I currently have a 4D matrix of images in the form height x width x RGB x imageNumber in which I would like to index with a 2D array without using a for loop. The 2D array is in the format of height x width with the values being the image number to index.

我已经将它与A for循环一起使用，但是由于速度原因，有一种方法可以不循环吗?我尝试过调整矩阵和索引数组的大小，但到目前为止还没有运气.

I've got it working with A for loop but due to speed is there a way to do it without looping? I've tried resizing the matrix and index array but no luck so far.

这是我正在工作的for循环(尽管在大图像上慢慢显示):

Here is the for loop I've got working (albeit slowly on large images):

for height = 1:h
    for width = 1:w
        imageIndex = index(height, width);
        imageOutput(height, width, :) = matrix4D(height, width, :, imageIndex);
    end
end

其中h和w是图像的高度和宽度尺寸.

where h and w are the height and width dimensions of the images.

谢谢！

推荐答案

这使用

This uses implicit expansion to build a linear index that produces the desired result:

matrix4D = rand(4,2,3,5); % example matrix
[h, w, c, n] = size(matrix4D); % sizes
index = randi(n,h,w); % example index
ind = reshape(1:h*w,h,w) + reshape((0:c-1)*h*w,1,1,[]) + (index-1)*h*w*c; % linear index
imageOutput = matrix4D(ind); % desired result

对于R2016b之前的Matlab版本，您需要使用 bsxfun 而不是隐式扩展:

For Matlab versions before R2016b you need to use bsxfun instead of implicit expansion:

ind = bsxfun(@plus, bsxfun(@plus, ...
    reshape(1:h*w,h,w), reshape((0:c-1)*h*w,1,1,[])), (index-1)*h*w*c); % linear index

这篇关于用2D索引矩阵索引4D数组的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！