rgb数据是怎么获得的

[breakmyhabits]

作者你好！
提供的数据集包含data_mat_rgb.mat，请问是怎么处理的？可否分享处理方法
我的邮箱是hnbcfzwx@163.com

[LiPang]

使用如下函数插值到RGB中心波段

def interpolate_hyperspectral_image(img, wave_lib, target_wavelengths):
    if len(target_wavelengths.shape) == 1:
        target_wavelengths = target_wavelengths[None]
    wave_lib_np = wave_lib
    wave_current_np = target_wavelengths
    f = interp1d(wave_lib_np, np.arange(len(wave_lib_np)))
    temp = [f(t) for t in wave_current_np]
    transform_matrix = []
    for t in temp:
        transform_matrix_curr = np.zeros((wave_current_np.shape[1],
                                          len(wave_lib_np)),
                                         dtype=np.float32)
        idx0 = np.arange(wave_current_np.shape[1], dtype=np.int64)
        idx1 = np.floor(t).astype(np.int64)
        idx2 = np.ceil(t).astype(np.int64)
        w1 = 1 - t % 1
        w2 = t % 1
        transform_matrix_curr[idx0, idx2] = w2
        transform_matrix_curr[idx0, idx1] = w1
        transform_matrix.append(transform_matrix_curr)
    transform_matrix = np.stack(transform_matrix)

    img = np.transpose(img, (2, 0, 1))
    img = np.einsum('s n, n w h -> s w h', transform_matrix[0], img)
    img = np.transpose(img, (1, 2, 0))
    return img

[linsenye]

使用如下函数插值到RGB中心波段

def interpolate_hyperspectral_image(img, wave_lib, target_wavelengths):
    if len(target_wavelengths.shape) == 1:
        target_wavelengths = target_wavelengths[None]
    wave_lib_np = wave_lib
    wave_current_np = target_wavelengths
    f = interp1d(wave_lib_np, np.arange(len(wave_lib_np)))
    temp = [f(t) for t in wave_current_np]
    transform_matrix = []
    for t in temp:
        transform_matrix_curr = np.zeros((wave_current_np.shape[1],
                                          len(wave_lib_np)),
                                         dtype=np.float32)
        idx0 = np.arange(wave_current_np.shape[1], dtype=np.int64)
        idx1 = np.floor(t).astype(np.int64)
        idx2 = np.ceil(t).astype(np.int64)
        w1 = 1 - t % 1
        w2 = t % 1
        transform_matrix_curr[idx0, idx2] = w2
        transform_matrix_curr[idx0, idx1] = w1
        transform_matrix.append(transform_matrix_curr)
    transform_matrix = np.stack(transform_matrix)

    img = np.transpose(img, (2, 0, 1))
    img = np.einsum('s n, n w h -> s w h', transform_matrix[0], img)
    img = np.transpose(img, (1, 2, 0))
    return img

Thank you for providing the function that interpolates hyperspectral into RGB . However, the details of the libraries and variables used in the code seems unclear. Could you provide an example (e.g., hsi_2_rgb.py/ipynb) that contains the complete extraction process?