在 Matlab 中读取大型结构化文件

发布于 2025-01-01 05:15:48 字数 538 浏览 0 评论 0原文

在matlab中，我想从一个结构化的、相当大的文件（大小：18+2048*2048字节）中读取，其中第18个字节分配给标题，其余的是像素图像强度。这里关心的是速度。正如您在下面的代码中看到的，对文件的多次访问大大降低了性能。您能建议任何更快地从文件中读取这些内容的方法吗？例如，读取缓冲区中的整个内容，我们可以使用“fread”函数从中读取内容。

fid = fopen(fileName, 'r', 'b');     % 'r' readonly and 'b' big endian
a= fread(fid,1,'uint16');
b1= fread(fid,1,'uint32');
b2= fread(fid,1,'uint32');
c1= fread(fid,1,'uint32');
c2= fread(fid,1,'uint32');
img=zeros (...
for i= (b1 + 1) : (b2 + 1)
   for j= (c1 + 1) : (c2 + 1)
      img(i, j) = fread(fid,1,'uint16');
   end
end

原文

In matlab I would like to read from a structured, and rather big, file (size: 18+2048*2048 bytes), where 18 fist bytes are assigned to the header and the rest are pixel image intensities.
The concern here is the speed. As you see in the code below multiple access to the file has considerably slowed down the performance.
Can you suggest any faster way of reading these contents from the file? e.g. reading entire thing in a buffer from which we can read using the "fread" function.

fid = fopen(fileName, 'r', 'b');     % 'r' readonly and 'b' big endian
a= fread(fid,1,'uint16');
b1= fread(fid,1,'uint32');
b2= fread(fid,1,'uint32');
c1= fread(fid,1,'uint32');
c2= fread(fid,1,'uint32');
img=zeros (...
for i= (b1 + 1) : (b2 + 1)
   for j= (c1 + 1) : (c2 + 1)
      img(i, j) = fread(fid,1,'uint16');
   end
end

分享到QQ

分享到微博

如果你对这篇内容有疑问，欢迎到本站社区发帖提问参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。

发布评论

需要登录才能够评论，你可以免费注册一个本站的账号。

殊姿 2025-01-08 05:15:48

只需计算 b1、b2、c1 和 c2 的总大小 n，并将其传递给单个 fread(fid,n,'uint16') 调用，而不是循环遍历它。然后对输出调用 reshape 使其成为二维数组。

回复收藏 0 原文

清风疏影 2025-01-08 05:15:48

您可以一次性读取图像数据。注意元素的顺序

img(b1+1:b2+1,c1+1:c2+1) = fread(fid, [c2-c1+1 b2-b1+1], 'uint16').';

另一件需要考虑的事情是将数据存储为 uint16 以避免整数到双精度转换：

img = zeros(s1, s2, 'uint16');
img(b1+1:b2+1,c1+1:c2+1) = fread(fid, [c2-c1+1 b2-b1+1], '*uint16').';

You can read the image data in one go. Note the order of the elements

img(b1+1:b2+1,c1+1:c2+1) = fread(fid, [c2-c1+1 b2-b1+1], 'uint16').';

One other thing to consider is storing the data as uint16 to avoid the integer to double conversion:

img = zeros(s1, s2, 'uint16');
img(b1+1:b2+1,c1+1:c2+1) = fread(fid, [c2-c1+1 b2-b1+1], '*uint16').';

回复收藏 0 原文

老街孤人 2025-01-08 05:15:48

看来你可以简单地一次读取整个图像，如下所示：

%... read header first so that file pointer is positioned at start of image data

M = b2 - b1 + 1; % width
N = c2 - c1 + 1; % height

[A, count] = fread(fid, [M, N], 'uint16');

if (count ~= M * N) % error check for number of elements read
    disp('Error reading image file');
end

img = zeros(...);
img(b1+1:b2+1, c1+1:c2+1) = A; % offset the image as necessary

It seems that you could simply read the whole image at once like so:

%... read header first so that file pointer is positioned at start of image data

M = b2 - b1 + 1; % width
N = c2 - c1 + 1; % height

[A, count] = fread(fid, [M, N], 'uint16');

if (count ~= M * N) % error check for number of elements read
    disp('Error reading image file');
end

img = zeros(...);
img(b1+1:b2+1, c1+1:c2+1) = A; % offset the image as necessary

回复收藏 0 原文

~没有更多了~