1-D wavelet transform on individual pixels of an image time series data

I am working on an imaging system, where I have a stack of images whose source of illumination is a sinusoidally varying source. Thus for any pixel p(x,y,t) in the image series , there is a sinusoidally varying intensity (starting from the first image to the last image and tracing a single pixel location). However the intensity is not stationary and also has harmonic contents. How may I get the component of each harmonic from the pixel location using wavelet transform?

Answer 1

[Gokul Nath S J]

Hi Faisal,

To extract the harmonics from a pixel location in an image series with sinusoidally varying intensity using wavelet transform, you can follow these general steps:

1. Load the image series into MATLAB as a 3D array, where each slice represents an image frame:

% Load the image series as a 3D array
image_series = load_image_series('path/to/images/*.png');   

The load_image_series function is a custom function that loads a series of images from a file path using the imread function.

2.For the pixel location of interest, extract the intensity values over time as a 1D array:

% Extract the intensity values for a pixel location as a 1D array
pixel_values = squeeze(image_series(x,y,:));

The squeeze function is used to remove singleton dimensions from the 3D array.

3. Decompose the pixel values into frequency components using wavelet transform:

% Perform wavelet decomposition on the pixel values 
[c,l] = wavedec(pixel_values, num_levels, wavelet_name);
% Extract the wavelet coefficients for a specific harmonic frequency
harmonic_coefficients = detcoef(c,l,harm_level);

4. Reconstruct the harmonic frequency component from the extracted coefficients:

% Reconstruct the pixel values corresponding to the harmonic frequency component
harmonic_pixel_values = wrcoef('d',c,l,wavelet_name,harm_level);

The wrcoef function reconstructs the signal from the wavelet coefficients using a specified wavelet type and decomposition level.

5. Repeat steps 3-4 for each harmonic component of interest.

By repeating these steps for different harmonic frequencies, you can extract the frequency components of interest from a pixel location in an image series with sinusoidally varying intensity using wavelet transform.

with regards,

Gokul Nath S J

[Faisal Sani Bala]

Hello Gokul,

Thanks for the reply, the load_image_series function is not available in the toolbox, do I write it from the scratch?

[Gokul Nath S J]

Hi Faisal,

Yes, you might need to write the function while incorporating a loop depending on the number of images.

[Faisal Sani Bala]

Hi Gokul,

I have created the function. thanks, but I discovered the signal multiresolution analyzer app, is it implementing the same thing ?