Spaces:
Sleeping
Sleeping
File size: 6,517 Bytes
925fbb1 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 | """
Sprite Image Enhancement Module
Uses Real-ESRGAN for high-quality upscaling
"""
import cv2
import numpy as np
import torch
from PIL import Image
import os
class SpriteProcessor:
"""Processor for enhancing sprite sheet images"""
def __init__(self):
self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
self.model = None
self._load_model()
def _load_model(self):
"""Load Real-ESRGAN model"""
try:
from realesrgan import RealESRGANer
from basicsr.archs.rrdbnet_arch import RRDBNet
# Create model
model = RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64,
num_block=23, num_grow_ch=32, scale=4)
# Initialize Real-ESRGAN
model_path = "weights/RealESRGAN_x4plus.pth"
if os.path.exists(model_path):
self.model = RealESRGANer(
scale=4,
model_path=model_path,
model=model,
tile=0,
pre_pad=0,
half=False,
device=self.device
)
else:
print("Warning: Real-ESRGAN model not found, using fallback enhancement")
self.model = None
except Exception as e:
print(f"Error loading Real-ESRGAN: {e}")
self.model = None
def enhance_image(self, image: np.ndarray, scale: int = 4) -> np.ndarray:
"""
Enhance image quality using Real-ESRGAN or fallback methods
Args:
image: Input image (BGR or BGRA)
scale: Upscaling factor (2 or 4)
Returns:
Enhanced image
"""
# Handle alpha channel
has_alpha = len(image.shape) == 3 and image.shape[2] == 4
if has_alpha:
# Separate alpha channel
bgr = image[:, :, :3]
alpha = image[:, :, 3]
else:
bgr = image
alpha = None
# Enhance RGB channels
if self.model is not None and scale > 1:
try:
# Convert BGR to RGB for the model
rgb = cv2.cvtColor(bgr, cv2.COLOR_BGR2RGB)
# Apply Real-ESRGAN
enhanced_rgb, _ = self.model.enhance(rgb, outscale=scale)
# Convert back to BGR
enhanced_bgr = cv2.cvtColor(enhanced_rgb, cv2.COLOR_RGB2BGR)
except Exception as e:
print(f"Real-ESRGAN failed, using fallback: {e}")
enhanced_bgr = self._fallback_enhance(bgr, scale)
else:
enhanced_bgr = self._fallback_enhance(bgr, scale)
# Enhance alpha channel if present
if alpha is not None and scale > 1:
enhanced_alpha = cv2.resize(alpha, None, fx=scale, fy=scale,
interpolation=cv2.INTER_NEAREST)
# Merge channels
enhanced_image = cv2.merge([enhanced_bgr, enhanced_alpha])
else:
enhanced_image = enhanced_bgr
return enhanced_image
def _fallback_enhance(self, image: np.ndarray, scale: int) -> np.ndarray:
"""
Fallback enhancement using OpenCV
Args:
image: Input BGR image
scale: Upscaling factor
Returns:
Enhanced image
"""
# Resize with high-quality interpolation
new_width = int(image.shape[1] * scale)
new_height = int(image.shape[0] * scale)
enhanced = cv2.resize(image, (new_width, new_height),
interpolation=cv2.INTER_CUBIC)
# Apply sharpening
kernel = np.array([[-1, -1, -1],
[-1, 9, -1],
[-1, -1, -1]])
enhanced = cv2.filter2D(enhanced, -1, kernel)
# Denoise
enhanced = cv2.fastNlMeansDenoisingColored(enhanced, None, 5, 5, 7, 21)
return enhanced
def sharpen_image(self, image: np.ndarray, strength: float = 1.0) -> np.ndarray:
"""
Apply sharpening filter
Args:
image: Input image
strength: Sharpening strength
Returns:
Sharpened image
"""
kernel = np.array([[-1, -1, -1],
[-1, 9, -1],
[-1, -1, -1]]) * strength
sharpened = cv2.filter2D(image, -1, kernel)
return sharpened
def remove_blur(self, image: np.ndarray) -> np.ndarray:
"""
Reduce blur using deconvolution
Args:
image: Input image
Returns:
Deblurred image
"""
# Create a point spread function (PSF)
psf_size = 5
psf = np.ones((psf_size, psf_size)) / (psf_size ** 2)
# Simple deconvolution (Wiener filter approximation)
result = image.copy()
for i in range(3): # For each channel
channel = image[:, :, i].astype(np.float32) / 255.0
# FFT
psf_fft = np.fft.fft2(psf, s=channel.shape)
channel_fft = np.fft.fft2(channel)
# Wiener deconvolution
K = 0.01 # Noise to signal ratio
deconv_fft = channel_fft * np.conj(psf_fft) / (np.abs(psf_fft) ** 2 + K)
# Inverse FFT
deconv = np.fft.ifft2(deconv_fft).real
# Clip and convert back
deconv = np.clip(deconv * 255, 0, 255).astype(np.uint8)
result[:, :, i] = deconv
return result
def enhance_contrast(self, image: np.ndarray) -> np.ndarray:
"""
Enhance contrast using CLAHE
Args:
image: Input image
Returns:
Contrast-enhanced image
"""
lab = cv2.cvtColor(image, cv2.COLOR_BGR2LAB)
l, a, b = cv2.split(lab)
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
l = clahe.apply(l)
enhanced = cv2.merge([l, a, b])
enhanced = cv2.cvtColor(enhanced, cv2.COLOR_LAB2BGR)
return enhanced
|