Score Dimensions

Image Quality

1. Sharpness

• Evaluation Method: Use image processing algorithms to calculate the sharpness of the image.
• Indicator: Laplacian Energy
• Reward Settings:
  • No Reward: Laplacian Energy < 40
  • Basic Reward: 40 ≤ Laplacian Energy < 60
  • Highest Reward: Laplacian Energy ≥ 60

import cv2
import numpy as np

def calculate_sharpness(image):
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    laplacian = cv2.Laplacian(gray, cv2.CV_64F)
    laplacian_energy = np.mean(np.square(laplacian))
    return laplacian_energy

def get_sharpness_reward(sharpness):
    if sharpness < 40:
        return 0
    elif 40 <= sharpness < 60:
        return 1
    else:
        return 1.2

# Example usage
image = cv2.imread('image.jpg')
sharpness = calculate_sharpness(image)
reward = get_sharpness_reward(sharpness)

1. Brightness

• Evaluation Method: Calculate the average brightness value of the image.
• Indicator: Average grayscale value (0-255)
• Reward Settings:
  • No Reward: Average grayscale value < 80 or > 200
  • Basic Reward: 80 ≤ Average grayscale value < 100 or 180 < Average grayscale value ≤ 200
  • Highest Reward: Average grayscale value 100 ≤ Average grayscale value ≤ 180

def calculate_brightness(image):
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    return np.mean(gray)

def get_brightness_reward(brightness):
    if brightness < 80 or brightness > 200:
        return 0
    elif 80 <= brightness < 100 or 180 < brightness <= 200:
        return 1
    else:
        return 1.2

# Example usage
brightness = calculate_brightness(image)
reward = get_brightness_reward(brightness)

1. Contrast

• Evaluation Method: Calculate the standard deviation of pixel grayscale values in the image.
• Indicator: Standard Deviation
• Reward Settings:
  • No Reward: Standard Deviation < 30
  • Basic Reward: 30 ≤ Standard Deviation < 50
  • Highest Reward: Standard Deviation ≥ 50

def calculate_contrast(image):
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    return np.std(gray)

def get_contrast_reward(contrast):
    if contrast < 30:
        return 0
    elif 30 <= contrast < 50:
        return 1
    else:
        return 1.2

# Example usage
contrast = calculate_contrast(image)
reward = get_contrast_reward(contrast)

1. Dynamic Range

• Evaluation Method: Calculate the ratio of the maximum brightness value to the minimum brightness value of the image.
• Indicator: Maximum brightness / Minimum brightness