def noise(src, p, p_chroma, level):
if np.random.uniform() < p:
with iproc.array_to_wand(src) as tmp:
tmp = _noise(tmp, p_chroma, level)
dst = iproc.wand_to_array(tmp)
return dst
else:
return src
def scale(src, filters, bmin, bmax, scale):
h, w = src.shape[:2]
blur = np.random.uniform(bmin, bmax)
rand = random.randint(0, len(filters) - 1)
with iproc.array_to_wand(src) as tmp:
tmp.resize(w // 2, h // 2, filters[rand], blur)
if scale:
tmp.resize(w, h, 'box')
dst = iproc.wand_to_array(tmp)
return dst
def noise_scale(src, filters, bmin, bmax, upscaling, p, p_chroma, level):
h, w = src.shape[:2]
blur = np.random.uniform(bmin, bmax)
rand = random.randint(0, len(filters) - 1)
with iproc.array_to_wand(src) as tmp:
tmp.resize(w // 2, h // 2, filters[rand], blur)
if np.random.uniform() < p:
tmp = _noise(tmp, p_chroma, level)
if upscaling:
tmp.resize(w, h, 'box')
dst = iproc.wand_to_array(tmp)
return dst
def scale(src, filters, bmin, bmax, upsampling):
# 'box', 'triangle', 'hermite', 'hanning', 'hamming', 'blackman',
# 'gaussian', 'quadratic', 'cubic', 'catrom', 'mitchell', 'lanczos',
# 'lanczos2', 'sinc'
h, w = src.shape[:2]
blur = np.random.uniform(bmin, bmax)
rand = random.randint(0, len(filters) - 1)
with iproc.array_to_wand(src) as tmp:
tmp.resize(w // 2, h // 2, filters[rand], blur)
if upsampling:
tmp.resize(w, h, 'box')
dst = iproc.wand_to_array(tmp)
return dst
def benchmark(cfg, models, images, sampling_factor, quality):
scores = []
for src in images:
dst = pairwise_transform.scale(np.array(src),
[cfg.downsampling_filter], 1, 1, False)
if quality != 100 or cfg.method != 'scale':
with iproc.array_to_wand(dst) as tmp:
tmp = iproc.jpeg(tmp, sampling_factor, quality)
dst = iproc.wand_to_array(tmp)
dst = Image.fromarray(dst)
if 'noise_scale' in models:
dst = upscale_image(cfg, dst, models['noise_scale'])
else:
if 'noise' in models:
dst = denoise_image(cfg, dst, models['noise'])
if 'scale' in models:
dst = upscale_image(cfg, dst, models['scale'])
score = iproc.clipped_psnr(np.array(dst), np.array(src), a_max=255)
scores.append(score)
return np.mean(scores), np.std(scores) / np.sqrt(len(scores))
