I have an array img_data of shape (x, x, n_channels) and I want to convolve / smooth along the axis=2. Specifically, I would like the output shape to be (x,x,n_channels//3), after convolving the frequency channels. I want to use a Lorentzian as a kernel for convolving. I build the kernel as: def lorentzian_kernel(x, center, gamma): norm = 1 / np.pi kernel = norm * (gamma / ((x - center) ** 2 + gamma ** 2)) return kernel So img_data is 3D array with n_channels depth and each channel is a square image. Each channel has a center frequency central_freq and FWHM which I shall use to make the kernels: gamma = FWHM / 2, center = central_freq and x = np.linspace(central_freq_channel1, central_freq_channeln, 1000) What operation should I use for smoothing along the channel axis? Do I have to do a dot product or convolution along axis=2 ? I am bit confused, so I would appreciate some comments. Thanks Continue reading...
