Question

我正在处理图像数据，我有240个图像，每个图像由5个通道贡献，大小231到384.这个矩阵现在大小（240,231,384,5），我希望这个（231 * 240,384， 5）。我不能在不扭曲数据的情况下“重塑”它。我该怎么做？

Answer 1

这是我对你要做的事情的猜测

test = [1:16]; % sample data
test1 = reshape(test, 4, 4); % rearrange to 4x4
% this has a vertical arrangement which is prob not what you have

[m, n] = size(test1); % get dims
r = 2; % horizontal width / numb of cols
% this prob closer to what you have
test2 = test1(:,1:r)'  % flip to horizontal order
test3 = reshape(test2, m*r, 1) % rearrange to vertical

Answer 2

给出一个矩阵：

n1 = 240; n2 = 231; n3 = 384; n4 = 5;
A = randn(n1, n2, n3, n4);

我想您正在寻求的解决方案是上面Luis所建议的：

B = reshape(A, n1 * n2, n3, n4);

或

C = reshape(permute(A, [2 1 3 4]), n1 * n2, n3, n4);

否则你必须更好地解释你的问题。

在matlab中连接部分多维矩阵

2 个答案: