I am working on a signal processing related problem. I have a dataset of >2000 EEG signals. Each EEG Signal is represented by a 2D Numpy array (19 x 30000). Each row of the array is one of the channels of the signal. What I have to do is to find the spectrograms on these individual channels (rows) and concatenate them vertically. Here is the code I wrote so far.
raw = np.load('class_1_ar/'+filename)
images = []
for i in range(19):
print(i,end=" ")
spec,freq,t,im = plt.specgram(raw[i],Fs=100,NFFT=100,noverlap=50)
plt.axis('off')
figure = plt.gcf()
figure.set_size_inches(12, 1)
figure.canvas.draw()
img = np.array(figure.canvas.buffer_rgba())
img = cv2.cvtColor(img, cv2.COLOR_RGBA2BGRA)
b = figure.axes[0].get_window_extent()
img = np.array(figure.canvas.buffer_rgba())
img = img[int(b.y0):int(b.y1),int(b.x0):int(b.x1),:]
img = cv2.cvtColor(img, cv2.COLOR_RGBA2BGRA)
images.append(img)
base = cv2.vconcat(images)
cv2.imwrite('class_1_sp/'+filename[:-4]+'.png',base)
c -= 1
print(c)
And here is my output:
However, the process is taking too much time to process. It took almost 8 hours for the first 200 samples to process.
My question is, What can I do to make it faster?
Like others have said, the overhead of going through matplotlib is likely slowing things down. It would be better to just compute (and not plot) the spectrogram with scipy.signal.spectrogram. This function directly returns the spectrogram as a 2D numpy array, so that you don't have the roundabout step of getting it out of the canvas. Note, that does mean you'll have to map the spectrogram output yourself to pixel intensities. In doing that, beware scipy.signal.spectrogram returns the spectrogram as powers, not decibels, so you probably want to do 10*np.log10(Sxx) to the result (see also scipy.signal.spectrogram compared to matplotlib.pyplot.specgram).
Plotting aside, the bottleneck operation in computing a spectrogram are the FFTs. Instead of using a transform size of 100 samples, 128 or some other power of 2 is more efficient. With scipy.signal.spectrogram this is done by setting nfft=128. Note, you can set nperseg=100 and nfft=128 so that 100 samples are still used for each segment, but zero-padded to 128 before doing the FFT. One other thought: if raw is 64-bit float, it may help to cast it to 32-bit: raw = np.load(...).astype(np.float32).
Related
I`m working on script/function for school project which will generate all 24-bit RGB colors image in matlab.
I wrote something like this but it is slow (and matlab don`t like me and crashes a lot). Last time before crash it was working on for 5 days.
Here is the code:
a = 1;
for r = 0:255
for g = 0:255
for b = 0:255
colors(a,:) = [r g b];
a = a + 1;
end
end
end
colors = reshape(colors, [4096, 4096, 3]);
colors = uint8(colors);
imshow(colors);
imwrite(colors, 'generated.png');
Is there any faster approach to do this?
Build the three columns separately using repmat/repelem and then concatenate them.
colors = [repelem((0:255).',256^2),...
repmat([repelem((0:255).',256) repmat((0:255).',256,1)],256,1)];
It is generally a good idea to preallocate large matrices to speed up code. With your current implementation, the size of colors grows one row every iteration which requires a lot memory allocation resources. Try to define your matrix with
colors = zeros(2^24, 3);
at the beginning of your code. To save memory and time, you can even define the matrix as uint8 from the beginning instead of converting it afterwards
colors = zeros(2^24, 3, 'uint8');
I am learning image analysis and trying to average set of color images and get standard deviation at each pixel
I have done this, but it is not by averaging RGB channels. (for ex rchannel = I(:,:,1))
filelist = dir('dir1/*.jpg');
ims = zeros(215, 300, 3);
for i=1:length(filelist)
imname = ['dir1/' filelist(i).name];
rgbim = im2double(imread(imname));
ims = ims + rgbim;
end
avgset1 = ims/length(filelist);
figure;
imshow(avgset1);
I am not sure if this is correct. I am confused as to how averaging images is useful.
Also, I couldn't get the matrix holding standard deviation.
Any help is appreciated.
If you are concerned about finding the mean RGB image, then your code is correct. What I like is that you converted the images using im2double before accumulating the mean and so you are making everything double precision. As what Parag said, finding the mean image is very useful especially in machine learning. It is common to find the mean image of a set of images before doing image classification as it allows the dynamic range of each pixel to be within a normalized range. This allows the training of the learning algorithm to converge quickly to the optimum solution and provide the best set of parameters to facilitate the best accuracy in classification.
If you want to find the mean RGB colour which is the average colour over all images, then no your code is not correct.
You have summed over all channels individually which is stored in sumrgbims, so the last step you need to do now take this image and sum over each channel individually. Two calls to sum in the first and second dimensions chained together will help. This will produce a 1 x 1 x 3 vector, so using squeeze after this to remove the singleton dimensions and get a 3 x 1 vector representing the mean RGB colour over all images is what you get.
Therefore:
mean_colour = squeeze(sum(sum(sumrgbims, 1), 2));
To address your second question, I'm assuming you want to find the standard deviation of each pixel value over all images. What you will have to do is accumulate the square of each image in addition to accumulating each image inside the loop. After that, you know that the standard deviation is the square root of the variance, and the variance is equal to the average sum of squares subtracted by the mean squared. We have the mean image, now you just have to square the mean image and subtract this with the average sum of squares. Just to be sure our math is right, supposing we have a signal X with a mean mu. Given that we have N values in our signal, the variance is thus equal to:
Source: Science Buddies
The standard deviation would simply be the square root of the above result. We would thus calculate this for each pixel independently. Therefore you can modify your loop to do that for you:
filelist = dir('set1/*.jpg');
sumrgbims = zeros(215, 300, 3);
sum2rgbims = sumrgbims; % New - for standard deviation
for i=1:length(filelist)
imname = ['set1/' filelist(i).name];
rgbim = im2double(imread(imname));
sumrgbims = sumrgbims + rgbim;
sum2rgbims = sum2rgbims + rgbim.^2; % New
end
rgbavgset1 = sumrgbims/length(filelist);
% New - find standard deviation
rgbstdset1 = ((sum2rgbims / length(filelist)) - rgbavgset.^2).^(0.5);
figure;
imshow(rgbavgset1, []);
% New - display standard deviation image
figure;
imshow(rgbstdset1, []);
Also to make sure, I've scaled the display of each imshow call so the smallest value gets mapped to 0 and the largest value gets mapped to 1. This does not change the actual contents of the images. This is just for display purposes.
I've found some methods to enlarge an image but there is no solution to shrink an image. I'm currently using the nearest neighbor method. How could I do this with bilinear interpolation without using the imresize function in MATLAB?
In your comments, you mentioned you wanted to resize an image using bilinear interpolation. Bear in mind that the bilinear interpolation algorithm is size independent. You can very well use the same algorithm for enlarging an image as well as shrinking an image. The right scale factors to sample the pixel locations are dependent on the output dimensions you specify. This doesn't change the core algorithm by the way.
Before I start with any code, I'm going to refer you to Richard Alan Peters' II digital image processing slides on interpolation, specifically slide #59. It has a great illustration as well as pseudocode on how to do bilinear interpolation that is MATLAB friendly. To be self-contained, I'm going to include his slide here so we can follow along and code it:
Please be advised that this only resamples the image. If you actually want to match MATLAB's output, you need to disable anti-aliasing.
MATLAB by default will perform anti-aliasing on the images to ensure the output looks visually pleasing. If you'd like to compare apples with apples, make sure you disable anti-aliasing when comparing between this implementation and MATLAB's imresize function.
Let's write a function that will do this for us. This function will take in an image (that is read in through imread) which can be either colour or grayscale, as well as an array of two elements - The image you want to resize and the output dimensions in a two-element array of the final resized image you want. The first element of this array will be the rows and the second element of this array will be the columns. We will simply go through this algorithm and calculate the output pixel colours / grayscale values using this pseudocode:
function [out] = bilinearInterpolation(im, out_dims)
%// Get some necessary variables first
in_rows = size(im,1);
in_cols = size(im,2);
out_rows = out_dims(1);
out_cols = out_dims(2);
%// Let S_R = R / R'
S_R = in_rows / out_rows;
%// Let S_C = C / C'
S_C = in_cols / out_cols;
%// Define grid of co-ordinates in our image
%// Generate (x,y) pairs for each point in our image
[cf, rf] = meshgrid(1 : out_cols, 1 : out_rows);
%// Let r_f = r'*S_R for r = 1,...,R'
%// Let c_f = c'*S_C for c = 1,...,C'
rf = rf * S_R;
cf = cf * S_C;
%// Let r = floor(rf) and c = floor(cf)
r = floor(rf);
c = floor(cf);
%// Any values out of range, cap
r(r < 1) = 1;
c(c < 1) = 1;
r(r > in_rows - 1) = in_rows - 1;
c(c > in_cols - 1) = in_cols - 1;
%// Let delta_R = rf - r and delta_C = cf - c
delta_R = rf - r;
delta_C = cf - c;
%// Final line of algorithm
%// Get column major indices for each point we wish
%// to access
in1_ind = sub2ind([in_rows, in_cols], r, c);
in2_ind = sub2ind([in_rows, in_cols], r+1,c);
in3_ind = sub2ind([in_rows, in_cols], r, c+1);
in4_ind = sub2ind([in_rows, in_cols], r+1, c+1);
%// Now interpolate
%// Go through each channel for the case of colour
%// Create output image that is the same class as input
out = zeros(out_rows, out_cols, size(im, 3));
out = cast(out, class(im));
for idx = 1 : size(im, 3)
chan = double(im(:,:,idx)); %// Get i'th channel
%// Interpolate the channel
tmp = chan(in1_ind).*(1 - delta_R).*(1 - delta_C) + ...
chan(in2_ind).*(delta_R).*(1 - delta_C) + ...
chan(in3_ind).*(1 - delta_R).*(delta_C) + ...
chan(in4_ind).*(delta_R).*(delta_C);
out(:,:,idx) = cast(tmp, class(im));
end
Take the above code, copy and paste it into a file called bilinearInterpolation.m and save it. Make sure you change your working directory where you've saved this file.
Except for sub2ind and perhaps meshgrid, everything seems to be in accordance with the algorithm. meshgrid is very easy to explain. All you're doing is specifying a 2D grid of (x,y) co-ordinates, where each location in your image has a pair of (x,y) or column and row co-ordinates. Creating a grid through meshgrid avoids any for loops as we will have generated all of the right pixel locations from the algorithm that we need before we continue.
How sub2ind works is that it takes in a row and column location in a 2D matrix (well... it can really be any amount of dimensions you want), and it outputs a single linear index. If you're not aware of how MATLAB indexes into matrices, there are two ways you can access an element in a matrix. You can use the row and column to get what you want, or you can use a column-major index. Take a look at this matrix example I have below:
A =
1 2 3 4 5
6 7 8 9 10
11 12 13 14 15
If we want to access the number 9, we can do A(2,4) which is what most people tend to default to. There is another way to access the number 9 using a single number, which is A(11)... now how is that the case? MATLAB lays out the memory of its matrices in column-major format. This means that if you were to take this matrix and stack all of its columns together in a single array, it would look like this:
A =
1
6
11
2
7
12
3
8
13
4
9
14
5
10
15
Now, if you want to access element number 9, you would need to access the 11th element of this array. Going back to the interpolation bit, sub2ind is crucial if you want to vectorize accessing the elements in your image to do the interpolation without doing any for loops. As such, if you look at the last line of the pseudocode, we want to access elements at r, c, r+1 and c+1. Note that all of these are 2D arrays, where each element in each of the matching locations in all of these arrays tell us the four pixels we need to sample from in order to produce the final output pixel. The output of sub2ind will also be 2D arrays of the same size as the output image. The key here is that each element of the 2D arrays of r, c, r+1, and c+1 will give us the column-major indices into the image that we want to access, and by throwing this as input into the image for indexing, we will exactly get the pixel locations that we want.
There are some important subtleties I'd like to add when implementing the algorithm:
You need to make sure that any indices to access the image when interpolating outside of the image are either set to 1 or the number of rows or columns to ensure you don't go out of bounds. Actually, if you extend to the right or below the image, you need to set this to one below the maximum as the interpolation requires that you are accessing pixels to one over to the right or below. This will make sure that you're still within bounds.
You also need to make sure that the output image is cast to the same class as the input image.
I ran through a for loop to interpolate each channel on its own. You could do this intelligently using bsxfun, but I decided to use a for loop for simplicity, and so that you are able to follow along with the algorithm.
As an example to show this works, let's use the onion.png image that is part of MATLAB's system path. The original dimensions of this image are 135 x 198. Let's interpolate this image by making it larger, going to 270 x 396 which is twice the size of the original image:
im = imread('onion.png');
out = bilinearInterpolation(im, [270 396]);
figure;
imshow(im);
figure;
imshow(out);
The above code will interpolate the image by increasing each dimension by twice as much, then show a figure with the original image and another figure with the scaled up image. This is what I get for both:
Similarly, let's shrink the image down by half as much:
im = imread('onion.png');
out = bilinearInterpolation(im, [68 99]);
figure;
imshow(im);
figure;
imshow(out);
Note that half of 135 is 67.5 for the rows, but I rounded up to 68. This is what I get:
One thing I've noticed in practice is that upsampling with bilinear has decent performance in comparison to other schemes like bicubic... or even Lanczos. However, when you're shrinking an image, because you're removing detail, nearest neighbour is very much sufficient. I find bilinear or bicubic to be overkill. I'm not sure about what your application is, but play around with the different interpolation algorithms and see what you like out of the results. Bicubic is another story, and I'll leave that to you as an exercise. Those slides I referred you to does have material on bicubic interpolation if you're interested.
Good luck!
I have a large set of microscopy images and each image has several hundreds of spots (ROIs). These spots are fixed in space. I want to extract each spot from each image and save into workspace so that I can analyze them further.
I have written a code myself and it working perfectly but its too slow. It takes around 250 sec to completely read out all the spots from every image.
The core of my code looks as following:
for s=1:NumberImages
im1=imread(fn(s,1).name);
im=im1-medfilt2(im1,[15,15]);
for i=1:length(p_g_x)
GreenROI(i,s)=double(sum(sum(im(round(p_g_y(i))+(-2:2), round(p_g_x(i))+(-2:2)))));
RedROI(i,s)=double(sum(sum(im(round(p_r_y(i))+(-2:2), round(p_r_x(i))+(-2:2)))));
end
end
As you can see from the code I am extracting 5x5 regions. Length of p_g_x is between 500-700.
Thanks for your input. I used profile viewer to figure out which function exactly is taking more time. It was median filter which is taking a lot of time (~90%).
Any suggestion to fast it up will be greatly appreciated.
thanks
Mahipal
Use Matlab's profiling tools!
profile on % Starts the profiler
% Run some code now.
profile viewer % Shows you how often each function was called, and
% where most time was spent. Try to start with the slowest part.
profile off % Resets the Profiler, so you can measure again.
Pre-allocate
Preallocate the output because you know the size and this way it is much faster. (Matlab told you this already!)
GreenROI = zeros(length(p_g_x), NumberImages); % And the same for RedROI.
Use convolution
Read about Matlab's conv2 code.
for s=1:NumberImages
im=imread(fn(s,1).name);
im=im-medfilt2(im,[15,15]);
% Pre-compute the sums first. This will only be faster for large p_g_x
roi_image = conv2(im, ones(5,5));
for i=1:length(p_g_x)
GreenROI(i,s)=roi_image(round(p_g_y(i)), round(p_g_x(i))); % You might have to offset the indices by 2, because of the convolution. Check that results are the same.
RedROI(i,s)=roi_image(round(p_r_y(i)), round(p_r_x(i)));
end
end
Matlab-ize the code
Now, that you've used convolution to get an image of sums over 5x5 windows (or you could've used #Shai's accumarray, same thing), you can speed things up further by not iterating through each element in p_g_x but use it as a vector straight away.
I leave that as an exercise for the reader. (convert p_g_x and p_g_y to indices using sub2ind, as a hint).
Update
Our answers, mine included, showed how premature optimisation is a bad thing. Without knowing, I assumed that your loop would take most of the time, but when you measured it (thanks!) it turns out that is not the problem. The bottleneck is medfilt2 the median filter, which takes 90% of the time. So you should address this first. (Note, that on my computer your original code is fast enough for my taste but it is still the median filter taking up most of the time.)
Looking at what the median filter operation does might help us figure out how to make it faster. Here is an image. On the left you see the original image. In the middle the median filter and on the right there is the result.
To me the result looks awfully similar to an edge detection result. (Mathematically this is no surprise.)
I would suggest you start experimenting with various edge detections. Have a look at Canny and Sobel. Or just use conv2(image, kernel_x) where kernel_x = [1, 2, 1; 0, 0, 0; -1, -2, -1] and the same but transposed for a kernel_y. You can find various edge detection options here: edge(im, option). I tried all options from {'sobel', 'canny', 'roberts', 'prewitt'}. Except for Canny, they all take about the same time as your median filter method. Canny is 4x slower, the rest (including the original) take 7.x seconds. All of this without a GPU. imgradient was 9 seconds.
So from this I would say that you can't get any faster. If you have a GPU and it works with Matlab, you could speed it up. Load your image as gpuArrays. There is an example on the medfilt2 documentation. You can still do minor speed ups but they can only amount to a 10% speed increase, so are hardly worthwile.
A few things you should do
Pre-allocate as suggested by Didac Perez.
Use profiler to see what exactly takes long in your code, is it the median filter? is it the indexing?
Assuming all images are of the same size, you can use accumarray and a fixed mask subs to quickly sum the values:
subs_g = zeros( h, w ); %// allocate mask for green
subs_r = zeros( h, w );
subs_g( sub2ind( [h w], round(p_g_y), round(p_g_x) ) = 1:numel(p_g_x); %//index each region
subs_g = conv2( subs_g, ones(5), 'same' );
subs_r( sub2ind( [h w], round(p_r_y), round(p_r_x) ) = 1:numel(p_r_x); %//index each region
subs_r = conv2( subs_r, ones(5), 'same' );
sel_g = subs_g > 0;
sel_r = subs_r > 0;
subs_g = subs_g(sel_g);
subs_r = subs_r(sel_r);
once these masks are fixed, you can process all images
%// pre-allocation goes here - I'll leave it to you
for s=1:NumberImages
im1=imread(fn(s,1).name);
im=double( im1-medfilt2(im1,[15,15]) );
accumarray( subs_g, im( sel_g ) ); % summing all the green ROIs
accumarray( subs_r, im( sel_r ) ); % summing all the green ROIs
end
First, preallocate your GreenROI and RedROI structures since you previously know the final size. Now, you are resizing them again and again in each iteration.
Secondly, I do recommend you to use "tic" and "toc" to investigate where is the problem, it will give you useful timings.
Vectorized code that operates on each image -
%// Pre-compute green and red indices to be used across all the images
r1 = round(bsxfun(#plus,permute(p_g_y,[3 2 1]),[-2:2]'));
c1 = round(bsxfun(#plus,permute(p_g_x,[3 2 1]),[-2:2]));
green_ind = reshape(bsxfun(#plus,(c1-1)*size(im,1),r1),[],numel(p_g_x));
r2 = round(bsxfun(#plus,permute(p_r_y,[3 2 1]),[-2:2]'));
c2 = round(bsxfun(#plus,permute(p_r_x,[3 2 1]),[-2:2]));
red_ind = reshape(bsxfun(#plus,(c2-1)*size(im,1),r2),[],numel(p_g_x));
for s=1:NumberImages
im1=imread(fn(s,1).name);
im=double(im1-medfilt2(im1,[15,15]));
GreenROI=sum(im(green_ind));
RedROI =sum(im(red_ind));
end
I've been performing a 2D mode filter on an RGB image by running medfilt2 independently on the R,G and B channels. However, splitting the RGB channels like this gives artifacts in the colouring. Is there a way to perform the 2D median filter while keeping RGB values 'together'?
Or, I could explain this more abstractly: Imagine I had a 2D matrix, where each value contained a pair of index coordinates (i.e. a cell matrix of 2X1 vectors). How would I go about performing a median filter on this?
Here's how I can do an independent mode filter (giving the artifacts):
r = colfilt(r0,[5 5],'sliding',#mode);
g = colfilt(g0,[5 5],'sliding',#mode);
b = colfilt(b0,[5 5],'sliding',#mode);
However colfilt won't work on a cell matrix.
Another approach could be to somehow combine my RGB channels into a single number and thus create a standard 2D matrix. Not sure how to implement this, though...
Any ideas?
Thanks for your help.
Cheers,
Hugh
EDIT:
OK, so problem solved. Here's how I did it.
I adapted my question so that I'm no longer dealing with (RGB) vectors, but (UV) vectors. Still essentially the same problem, except that my vectors are 2D not 3D.
So firstly I load the individual U and V channels, arrange them each into a 1D list, then combine them, so I essentially have a list of vectors. Then I reduce it to just those which are unique. Then, I assign each pixel in my matrix the value of the index of that unique vector. After this I can do the mode filter. Then I basically do the reverse, in that I go through the filtered image pixelwise, and read the value at each pixel (i.e. an index in my list), and find the unique vector associated with that index and insert it at that pixel.
% Create index list
img_u = img_iuv(:,:,2);
img_v = img_iuv(:,:,3);
coordlist = unique(cat(2,img_u(:),img_v(:)),'rows');
% Create a 2D matrix of indices
img_idx = zeros(size(img_iuv,1),size(img_iuv,2),2);
for y = 1:length(Y)
for x = 1:length(X)
coords = squeeze(img_iuv(x,y,2:3))';
[~,idx] = ismember(coords,coordlist,'rows');
img_idx(x,y) = idx;
end
end
% Apply the mode filter
img_idx = colfilt(img_idx,[n,n],'sliding',#mode);
% Re-construct the original image using the filtered data
for y = 1:length(Y)
for x = 1:length(X)
idx = img_idx(x,y);
try
coords = coordlist(idx,:);
end
img_iuv(x,y,2:3) = coords(:);
end
end
Not pretty but it gets the job done. I suppose this approach would also work for RGB images, or other similar situations.
Cheers,
Hugh
I don't see how you can define the median of a vector variable. You probably need to reduce the R,G,B components to a single value and then compunte the median on that value. Why not use the intensity level as that single value? You could do it easily with rgb2gray.