Files
ProjectBabble/image_quant.py
T

48 lines
1.3 KiB
Python

import cv2
import numpy as np
import struct
import zstd
def quantize_grayscale(image: np.ndarray, num_bits: int) -> np.ndarray:
"""
Quantizes a grayscale image to a specified number of bits.
Args:
image (np.ndarray): Grayscale image (values between 0 and 255).
num_bits (int): Number of bits for quantization (1-8).
Returns:
np.ndarray: Quantized grayscale image.
"""
if not (1 <= num_bits <= 8):
raise ValueError("num_bits must be between 1 and 8.")
# Calculate the number of levels
num_levels = 2 ** num_bits
# Normalize the image to the range [0, num_levels - 1]
quantized = (image / 255.0) * (num_levels - 1)
# Round to nearest integer and convert back to integers
quantized = np.round(quantized).astype(np.uint8)
# Scale back to the full grayscale range [0, 255]
quantized = (quantized * (255 // (num_levels - 1))).astype(np.uint8)
return quantized
image = cv2.imread("0001.png")
image = cv2.resize(image, (194, 194))
image = cv2.GaussianBlur(image, 2)
image = quantize_grayscale(image, 4)
data = image.tobytes()
frame_size = len(data)
compressed_size = len(zstd.compress(data, 5))
est_fps = 225/(compressed_size / 1000)
print(frame_size)
print(compressed_size)
print(est_fps)
cv2.imshow("yeah", image)
cv2.waitKey(0)