Getting black images while applying discrete cosine transform (DCT) on image

Looks like this

def perform_dct(image_path):
    # Load the image in grayscale
    img = cv2.imread(image_path, cv2.IMREAD_GRAYSCALE)
    if img is None:
        raise ValueError(f"Image at {image_path} could not be loaded.")
    # Resize the image to 224x224 (size expected by ViT model)


    img = cv2.resize(img, (224, 224))
    # Perform DCT
    dct_img = cv2.dct(np.float32(img))

       # Normalize the DCT image to the range [0, 1]
    dct_img_min = np.min(dct_img)
    dct_img_max = np.max(dct_img)
    dct_img_normalized = (dct_img - dct_img_min) / (dct_img_max - dct_img_min)
    
    # Convert to 3-channel RGB image
    dct_img_normalized = (dct_img_normalized * 255).astype(np.uint8)
    dct_img_rgb = cv2.merge([dct_img_normalized, dct_img_normalized, dct_img_normalized])

    plt.imshow(dct_img_rgb)
    plt.title("DCT of Imagesss")
    plt.show()

    return dct_img_rgb

Trying to get discrete cosine transform (DCT) image but getting whole black image , I am not able to find error in my code . If you can Please help me

Solution

I think that the result is expectable. When you run dct you may get in the output very large coefficients. Consider checking the distributions of the quantiles in dct_img array:

np.quantile(dct_img,[0,0.01,0.99,1])

dct_img values range:

min_value = -2537.42236328
1% quantile = -71.6647377
99% quantile = 69.59934616 
max_value = 36856.75390625

Given such high maximum after the normalisation (with min/max) 99% of values will be mapped to zeros. And this results in seeing only black pixels on the image.

I'm not sure why you want to plot dct_image array. Basically plotting it's coefficients is not required step for further analysis. But if you really would like to visualize it, I would suggest first scaling it's values by applying log:

dct_img = cv2.dct(np.float32(img))

# scale all dct values to reduce difference between extereme `max` and `min` values
scaled_dct = np.sign(dct_img) * np.log(1+np.abs(dct_img))

# normalize the scaled data
scaled_dct_normalized = normalize(scaled_dct)

# plot the output
plt.imshow(scaled_dct_normalized, cmap='gray', vmin=0, vmax=255)
plt.title("scalled DCT of Imagesss")
plt.show()