def __getitem__(self, idx):
idx = idx % len(self.img_list)
hr_path = osp.join(self.dataroot_hr, self.img_list[idx])
base, ext = osp.splitext(self.img_list[idx])
lr_basename = base + 'x{}'.format(self.scale) + ext
lr_path = osp.join(self.dataroot_lr, lr_basename)
hr = self.load_img(hr_path)
lr = self.load_img(lr_path)
if self.noise is not None:
lr = self.add_noise(lr, self.noise['type'], self.noise['value'])
if self.train_Y:
lr = rgb2ycbcr(lr)[:, :, np.newaxis]
hr = rgb2ycbcr(hr)[:, :, np.newaxis]
data = {}
if self.split == 'train':
lr_patch, hr_patch = self.get_patch(lr, hr, self.ps, self.scale)
lr, hr = self.augment(lr_patch, hr_patch, self.use_flip,
self.use_rot)
lr, hr = self.np2tensor(lr), self.np2tensor(hr)
data['LR'] = lr
data['HR'] = hr
return data
def __getitem__(self, index):
input_ = self.input[index, :, :, :]
target_ = self.target[index, :, :, :]
subim_in, subim_tar = get_patch(input_, target_, self.patch_size,
self.sr_factor)
if not self.rgb:
subim_in = utils.rgb2ycbcr(subim_in)
subim_tar = utils.rgb2ycbcr(subim_tar)
subim_in = np.expand_dims(subim_in[:, :, 0], 2)
subim_tar = np.expand_dims(subim_tar[:, :, 0], 2)
if self.input_up:
subim_bic = imresize(subim_in, scalar_scale=self.sr_factor)
subim_in = utils.np2tensor(subim_in, self.rgb_range)
subim_tar = utils.np2tensor(subim_tar, self.rgb_range)
subim_bic = utils.np2tensor(subim_bic, self.rgb_range)
return {
'input': subim_in,
'target': subim_tar,
'input_up': subim_bic
}
subim_in = utils.np2tensor(subim_in, self.rgb_range)
subim_tar = utils.np2tensor(subim_tar, self.rgb_range)
return {'input': subim_in, 'target': subim_tar}
def comput_PSNR_SSIM(self, pred, gt, shave_border=0):
if isinstance(pred, torch.Tensor):
pred = tensor2np(pred, self.opt.rgb_range)
pred = pred.astype(np.float32)
if isinstance(gt, torch.Tensor):
gt = tensor2np(gt, self.opt.rgb_range)
gt = gt.astype(np.float32)
height, width = pred.shape[:2]
pred = pred[shave_border:height - shave_border,
shave_border:width - shave_border]
gt = gt[shave_border:height - shave_border,
shave_border:width - shave_border]
if pred.shape[2] == 3 and gt.shape[2] == 3:
pred_y = rgb2ycbcr(pred)[:, :, 0]
gt_y = rgb2ycbcr(gt)[:, :, 0]
elif pred.shape[2] == 1 and gt.shape[2] == 1:
pred_y = pred[:, :, 0]
gt_y = gt[:, :, 0]
else:
raise ValueError('Input or output channel is not 1 or 3!')
psnr_ = calc_PSNR(pred_y, gt_y)
ssim_ = calc_ssim(pred_y, gt_y)
return psnr_, ssim_
def gen_tfrecords(self, save_dir='DIV2K/tfrecords', tfrecord_num=10):
file_num = tfrecord_num
sample_num = 0
if not os.path.exists(save_dir):
os.makedirs(save_dir)
fs = []
for i in range(file_num):
fs.append(
tf.python_io.TFRecordWriter(
os.path.join(save_dir, 'data%d.tfrecords' % i)))
img_paths = sorted(glob.glob(os.path.join(self.data_dir, '*.png')))
for img_path in img_paths:
print('processing %s' % img_path)
img = misc.imread(img_path)
y = utils.img_to_uint8(utils.rgb2ycbcr(img)[:, :, 0])
height, width = y.shape
p = self.patch_size
step = p // 3 * 2
for h in range(0, height - p + 1, step):
for w in range(0, width - p + 1, step):
gt = y[h:h + p, w:w + p]
assert gt.shape[:] == (p, p)
assert gt.dtype == np.uint8
example = tf.train.Example(features=tf.train.Features(
feature={'gt': _bytes_feature(gt.tostring())}))
fs[sample_num % file_num].write(
example.SerializeToString())
sample_num += 1
print('example number: %d' % sample_num)
np.savetxt(os.path.join(save_dir, 'sample_num.txt'),
np.asarray([sample_num]), '%d')
for f in fs:
f.close()
def upscale(self, im_l, s):
"""
% im_l: LR image, float np array in [0, 255]
% im_h: HR image, float np array in [0, 255]
"""
im_l = im_l/255.0
if len(im_l.shape)==3 and im_l.shape[2]==3:
im_l_ycbcr = utils.rgb2ycbcr(im_l)
else:
im_l_ycbcr = np.zeros([im_l.shape[0], im_l.shape[1], 3])
im_l_ycbcr[:, :, 0] = im_l
im_l_ycbcr[:, :, 1] = im_l
im_l_ycbcr[:, :, 2] = im_l
im_l_y = im_l_ycbcr[:, :, 0]*255 #[16 235]
im_h_y = self.upscale_alg(im_l_y, s)
# recover color
if len(im_l.shape)==3:
im_ycbcr = utils.imresize(im_l_ycbcr, s);
im_ycbcr[:, :, 0] = im_h_y/255.0; #[16/255 235/255]
im_h = utils.ycbcr2rgb(im_ycbcr)*255.0
else:
im_h = im_h_y
im_h = np.clip(im_h, 0, 255)
im_h_y = np.clip(im_h_y, 0, 255)
return im_h,im_h_y
def upscale(self, im_l, s):
"""
% im_l: LR image, float np array in [0, 255]
% im_h: HR image, float np array in [0, 255]
"""
im_l = im_l/255.0
if len(im_l.shape)==3 and im_l.shape[2]==3:
im_l_ycbcr = utils.rgb2ycbcr(im_l)
else:
im_l_ycbcr = np.zeros([im_l.shape[0], im_l.shape[1], 3])
im_l_ycbcr[:, :, 0] = im_l
im_l_ycbcr[:, :, 1] = im_l
im_l_ycbcr[:, :, 2] = im_l
im_l_y = im_l_ycbcr[:, :, 0]*255 #[16 235]
im_h_y = self.upscale_alg(im_l_y, s)
# recover color
#print 'recover color...'
if len(im_l.shape)==3:
im_ycbcr = utils.imresize(im_l_ycbcr, s);
im_ycbcr[:, :, 0] = im_h_y/255.0; #[16/255 235/255]
im_h = utils.ycbcr2rgb(im_ycbcr)*255.0
else:
im_h = im_h_y
#print 'clip...'
im_h = np.clip(im_h, 0, 255)
im_h_y = np.clip(im_h_y, 0, 255)
return im_h,im_h_y
def gen_generator(self):
img_paths = sorted(glob.glob(os.path.join(self.data_dir, '*.png')))
one_img_patch_num = self.one_img_patch_num
p = self.patch_size
scale = self.scale
for img_path in img_paths:
img = misc.imread(img_path)
height, width, _ = img.shape
for i in range(one_img_patch_num):
h = np.random.randint(height - p + 1)
w = np.random.randint(width - p + 1)
patch = img[h:h + p, w:w + p]
gt = utils.rgb2ycbcr(patch)[:, :, 0]
gt = np.float32(gt) / 255.0
c1 = np.random.rand() #0~1均匀分布
c2 = np.random.rand()
if c1 < 0.5:
gt = gt[::-1, :]
if c2 < 0.5:
gt = gt[:, ::-1]
lr = misc.imresize(gt, 1.0 / scale, 'bicubic', 'F')
bic = misc.imresize(lr, scale / 1.0, 'bicubic', 'F')
yield lr, bic, gt
def __getitem__(self, index):
path = self.test_path[index]
images = {}
images.update({'name': path['save']})
input_ = cv2.imread(path['input'])
input_ = cv2.cvtColor(input_, cv2.COLOR_BGR2RGB)
target_ = cv2.imread(path['target'])
target_ = utils.modcrop(target_, self.sr_factor)
target_ = cv2.cvtColor(target_, cv2.COLOR_BGR2RGB)
if not self.rgb:
input_out = np.copy(input_)
input_out = utils.np2tensor(input_out, self.rgb_range)
# print(input_out)
input_ = utils.rgb2ycbcr(input_)
input_cbcr = input_[:, :, 1:]
input_ = np.expand_dims(input_[:, :, 0], 2)
input_cbcr = utils.np2tensor(input_cbcr, self.rgb_range)
images.update({'input_cbcr': input_cbcr, 'input_rgb': input_out})
if self.target_down:
target_down = imresize(target_, scalar_scale=1 / self.sr_factor)
target_down = utils.np2tensor(target_down, self.rgb_range)
images.update({'target_down': target_down})
input_ = utils.np2tensor(input_, self.rgb_range)
target_ = utils.np2tensor(target_, self.rgb_range)
images.update({'input': input_, 'target': target_})
return images
def __getitem__(self, index):
path = self.train_path[index]
images = {}
if self.npy_reader:
input_ = np.load(path['input'], allow_pickle=False)
target_ = np.load(path['target'], allow_pickle=False)
target_ = utils.modcrop(target_, self.sr_factor)
else:
input_ = cv2.imread(path['input'])
input_ = cv2.cvtColor(input_, cv2.COLOR_BGR2RGB)
target_ = cv2.imread(path['target'])
target_ = utils.modcrop(target_, self.sr_factor)
target_ = cv2.cvtColor(target_, cv2.COLOR_BGR2RGB)
# for i in range(10):
# subim_in, subim_tar = get_patch(input_, target_, self.patch_size, self.sr_factor)
# win_mean = ndimage.uniform_filter(subim_in[:, :, 0], (5, 5))
# win_sqr_mean = ndimage.uniform_filter(subim_in[:, :, 0]**2, (5, 5))
# win_var = win_sqr_mean - win_mean**2
#
# if np.sum(win_var) / (win_var.shape[0]*win_var.shape[1]) > 30:
# break
subim_in, subim_tar = get_patch(input_, target_, self.patch_size,
self.sr_factor)
if not self.rgb:
subim_in = utils.rgb2ycbcr(subim_in)
subim_tar = utils.rgb2ycbcr(subim_tar)
subim_in = np.expand_dims(subim_in[:, :, 0], 2)
subim_tar = np.expand_dims(subim_tar[:, :, 0], 2)
if self.target_down:
subim_target_down = imresize(subim_tar,
scalar_scale=1 / self.sr_factor)
subim_target_down = utils.np2tensor(subim_target_down,
self.rgb_range)
images.update({'target_down': subim_target_down})
subim_in = utils.np2tensor(subim_in, self.rgb_range)
subim_tar = utils.np2tensor(subim_tar, self.rgb_range)
images.update({'input': subim_in, 'target': subim_tar})
return images
def __getitem__(self, index):
input_ = np.load(self.input_path[index])
target_ = np.load(self.target_path[index])
if not self.rgb:
input_ = utils.rgb2ycbcr(input_)
target_ = utils.rgb2ycbcr(target_)
if self.input_up:
input_bic = imresize(input_, scalar_scale=self.sr_factor).round()
input_ = utils.np2tensor(input_, self.rgb_range)
target_ = utils.np2tensor(target_, self.rgb_range)
input_bic_ = utils.np2tensor(input_bic, self.rgb_range)
return {'input': input_, 'target': target_, 'input_up': input_bic_}
input_ = utils.np2tensor(input_, self.rgb_range)
target_ = utils.np2tensor(target_, self.rgb_range)
return {'input': input_, 'target': target_}
def evaluate(self):
opt = self.opt
self.net.eval()
psnr = 0
for i, inputs in enumerate(self.test_loader):
HR = inputs[0].to(self.dev)
LR = inputs[1].to(self.dev)
ORI_LR = LR.clone().detach()
# match the resolution of (LR, HR) due to CutBlur
if HR.size() != LR.size():
scale = HR.size(2) // LR.size(2)
LR = F.interpolate(LR, scale_factor=scale, mode="nearest")
SR = self.net(LR)
if isinstance(SR, (list, tuple)):
SR = SR[-1]
SR = SR.detach()
# iter over batch
for i in range(HR.size(0)):
hr = HR[i].clamp(0,
255).round().cpu().byte().permute(1, 2,
0).numpy()
sr = SR[i].clamp(0,
255).round().cpu().byte().permute(1, 2,
0).numpy()
hr = hr[opt.crop:-opt.crop, opt.crop:-opt.crop, :]
sr = sr[opt.crop:-opt.crop, opt.crop:-opt.crop, :]
if opt.eval_y_only:
hr = utils.rgb2ycbcr(hr)
sr = utils.rgb2ycbcr(sr)
psnr += utils.calculate_psnr(hr, sr)
if opt.save_result:
save_root = os.path.join(opt.save_root, opt.dataset)
save_path = os.path.join(save_root, "{:04d}.png".format(i + 1))
utils.save_batch_hr_lr(HR, SR, ORI_LR, save_path)
# io.imsave(save_path, SR)
self.net.train()
return psnr / len(self.test_loader.dataset)
def __getitem__(self, index):
img_path = self.images_path[index]
target_ = cv2.imread(img_path)
input_ = imresize(target_, scalar_scale=1 / self.sr_factor)
if not self.rgb:
input_ = utils.rgb2ycbcr(input_)
target_ = utils.rgb2ycbcr(target_)
if self.input_up:
input_bic = imresize(input_, scalar_scale=self.sr_factor).round()
input_ = utils.np2tensor(input_, self.rgb_range)
target_ = utils.np2tensor(target_, self.rgb_range)
input_bic_ = utils.np2tensor(input_bic, self.rgb_range)
return {'input': input_, 'target': target_, 'input_up': input_bic_}
input_ = utils.np2tensor(input_, self.rgb_range)
target_ = utils.np2tensor(target_, self.rgb_range)
return {'input': input_, 'target': target_}
def __getitem__(self, idx):
hr_path = osp.join(self.dataroot_hr, self.img_list[idx])
lr_path = osp.join(self.dataroot_lr, self.img_list[idx])
hr = self.load_img(hr_path)
lr = self.load_img(lr_path)
if self.noise is not None:
lr = self.add_noise(lr, self.noise['type'], self.noise['value'])
if self.train_Y:
lr = rgb2ycbcr(lr)
hr = rgb2ycbcr(hr)
data = {}
if self.split == 'train':
lr_patch, hr_patch = self.get_patch(lr, hr, self.ps, self.scale)
lr, hr = self.augment(lr_patch, hr_patch, self.use_flip, self.use_rot)
lr ,hr = self.np2tensor(lr), self.np2tensor(hr)
data['LR'] = lr
data['HR'] = hr
return data
def evaluate(self):
opt = self.opt
self.net.eval()
if opt.save_result:
save_root = os.path.join(opt.save_root, opt.dataset)
os.makedirs(save_root, exist_ok=True)
psnr = 0
for i, inputs in enumerate(self.test_loader):
HR = inputs[0].to(self.dev)
LR = inputs[1].to(self.dev)
# match the resolution of (LR, HR) due to CutBlur
if HR.size() != LR.size():
scale = HR.size(2) // LR.size(2)
LR = F.interpolate(LR, scale_factor=scale, mode="nearest")
SR = self.net(LR).detach()
HR = HR[0].clamp(0, 255).round().cpu().byte().permute(1, 2,
0).numpy()
SR = SR[0].clamp(0, 255).round().cpu().byte().permute(1, 2,
0).numpy()
if opt.save_result:
save_path = os.path.join(save_root, "{:04d}.png".format(i + 1))
io.imsave(save_path, SR)
HR = HR[opt.crop:-opt.crop, opt.crop:-opt.crop, :]
SR = SR[opt.crop:-opt.crop, opt.crop:-opt.crop, :]
if opt.eval_y_only:
HR = utils.rgb2ycbcr(HR)
SR = utils.rgb2ycbcr(SR)
psnr += utils.calculate_psnr(HR, SR)
self.net.train()
return psnr / len(self.test_loader)
def evalimg(im_h_y, im_gt, shave=0):
if len(im_gt.shape)==3:
im_gt_ycbcr = utils.rgb2ycbcr(im_gt/255.0)*255.0
im_gt_y = im_gt_ycbcr[:, :, 0]
else:
im_gt_y = im_gt
diff = im_h_y.astype(np.uint8).astype(np.float32) - im_gt_y.astype(np.uint8).astype(np.float32)
#diff = im_h_y - im_gt_y
if shave>0:
diff = utils.shave(diff, [shave, shave])
res = {}
res['rmse'] = np.sqrt((diff**2).mean())
res['psnr'] = 20*np.log10(255.0/res['rmse'])
return res
def evalimg(im_h_y, im_gt, shave=0):
if len(im_gt.shape) == 3:
im_gt_ycbcr = utils.rgb2ycbcr(im_gt / 255.0) * 255.0
im_gt_y = im_gt_ycbcr[:, :, 0]
else:
im_gt_y = im_gt
diff = im_h_y.astype(np.uint8).astype(np.float32) - im_gt_y.astype(
np.uint8).astype(np.float32)
#diff = im_h_y - im_gt_y
if shave > 0:
diff = utils.shave(diff, [shave, shave])
res = {}
res['rmse'] = np.sqrt((diff**2).mean())
res['psnr'] = 20 * np.log10(255.0 / res['rmse'])
return res
def evalimg(im_h_y, im_gt, shave=0):
if len(im_gt.shape) == 3:
im_gt_ycbcr = utils.rgb2ycbcr(im_gt / 255.0) * 255.0
im_gt_y = im_gt_ycbcr[:, :, 0]
else:
im_gt_y = im_gt
im_h_y_uint8 = np.rint(np.clip(im_h_y, 0, 255))
im_gt_y_uint8 = np.rint(np.clip(im_gt_y, 0, 255))
diff = im_h_y_uint8 - im_gt_y_uint8
#diff = im_h_y - im_gt_y
if shave > 0:
diff = utils.shave(diff, [shave, shave])
res = {}
res['rmse'] = np.sqrt((diff**2).mean())
res['psnr'] = 20 * np.log10(255.0 / res['rmse'])
return res
def slice_reconstruction(size, slice, ang_tar):
# ---------------- Model -------------------- #
global slice_y
with sess.as_default():
if FLAG_RGB:
# slice_ycbcr = utils.rgb2ycbcr(slice)
slice = np.transpose(slice, (1, 0, 2, 3))
slice = np.expand_dims(slice, axis=0)
slice_y = slice[:, :, :, :, 0:1]
slice_cb = slice[:, :, :, :, 1:2]
slice_cr = slice[:, :, :, :, 2:3]
slice_y = sess.run(y_out, feed_dict={x: slice_y})
slice_cb = sess.run(y_out, feed_dict={x: slice_cb})
slice_cr = sess.run(y_out, feed_dict={x: slice_cr})
slice_ycbcr = np.concatenate((slice_y, slice_cb, slice_cr),
axis=-1)
slice_ycbcr = np.transpose(slice_ycbcr[0, :, :, :, :],
(1, 0, 2, 3))
slice_ycbcr = tf.convert_to_tensor(slice_ycbcr)
slice_ycbcr = tf.image.resize_bicubic(slice_ycbcr,
[ang_tar, size])
slice = sess.run(slice_ycbcr)
# slice = utils.ycbcr2rgb(slice_ycbcr)
else:
slice_ycbcr = utils.rgb2ycbcr(slice)
slice_y = np.transpose(slice_ycbcr[:, :, :, 0:1], (1, 0, 2, 3))
slice_ycbcr = tf.convert_to_tensor(slice_ycbcr)
slice_ycbcr = tf.image.resize_bicubic(slice_ycbcr,
[ang_tar, size])
slice_ycbcr = sess.run(slice_ycbcr)
slice_y = np.expand_dims(slice_y, axis=0)
slice_y = sess.run(y_out, feed_dict={x: slice_y})
slice_y = tf.convert_to_tensor(
np.transpose(slice_y[0], (1, 0, 2, 3)))
slice_y = tf.image.resize_bicubic(slice_y, [ang_tar, size])
slice_ycbcr[:, :, :, 0:1] = sess.run(slice_y)
slice = utils.ycbcr2rgb(slice_ycbcr)
slice = np.minimum(np.maximum(slice, 0), 1)
return slice
def evaluate(model, update_step, writer, bucket, engine):
device = torch.device('cuda:0')
model.eval()
eval_paths = [os.path.join(args.eval_path, v) for v in ['Set14', 'Set5']]
metrics_list = []
unnorm = UnNormalize(0.5, 0.5)
for eval_path in eval_paths:
eval_name = os.path.basename(eval_path)
HQ_path = os.path.join(eval_path, eval_name) + '.lmdb'
LQ_path = os.path.join(eval_path, eval_name) + '_LQ.lmdb'
LQ_r_path = os.path.join(eval_path, eval_name) + '_LQ_restored.lmdb'
eval_set = ValDataset(HQ_path, LQ_path, LQ_r_path, args.scale)
eval_loader = DataLoader(
eval_set, batch_size=1, shuffle=False, num_workers=4)
psrn_rgb = 0.0
psrn_y = 0.0
ssim_rgb = 0.0
ssim_y = 0.0
for i, data_dict in enumerate(eval_loader):
img_HQ = data_dict['img_GT']
img_LQ = data_dict['img_LQ'].to(device)
img_LQ_r = data_dict['img_LQ_r']
with torch.no_grad():
# SR image range [-1, 1]
img_SR = model(img_LQ)
# SR image range [0, 1]
img_SR = unnorm(img_SR)
if i == 0:
imgs = torch.cat([img_HQ, img_SR.detach().cpu(), img_LQ_r], dim=0)
grid = vutils.make_grid(imgs, nrow=3, normalize=False)
tmp_image = T.ToPILImage()(grid)
tmp_image.save('images/tmp_image.png')
upload_to_cloud(bucket, 'images/tmp_image.png',
'odesr01_04/image_progress/{}/gen_step_{}'.
format(eval_name, update_step * args.update_freq))
if eval_name == 'Set5':
writer.add_image('Set5', grid, update_step)
crop_size = args.scale
img_HQ_rgb = img_HQ[0].permute(2, 1, 0).cpu(). \
numpy()[crop_size:-crop_size, crop_size:-crop_size, :]
img_SR_rgb = img_SR[0].permute(2, 1, 0).detach().cpu(). \
numpy()[crop_size:-crop_size, crop_size:-crop_size, :]
img_HQ_y = rgb2ycbcr(img_HQ_rgb)
img_SR_y = rgb2ycbcr(img_SR_rgb)
psrn_rgb += calculate_psnr(img_HQ_rgb * 255, img_SR_rgb * 255)
psrn_y += calculate_psnr(img_HQ_y * 255, img_SR_y * 255)
ssim_rgb += calculate_ssim(img_HQ_rgb * 255, img_SR_rgb * 255)
ssim_y += calculate_ssim(img_HQ_y * 255, img_SR_y * 255)
psrn_rgb = psrn_rgb / len(eval_loader.dataset)
psrn_y = psrn_y / len(eval_loader.dataset)
ssim_rgb = ssim_rgb / len(eval_loader.dataset)
ssim_y = ssim_y / len(eval_loader.dataset)
metrics_list.extend([psrn_rgb, psrn_y, ssim_rgb, ssim_y])
if eval_name == 'Set5':
writer.add_scalar('psrn_rgb', psrn_rgb, update_step)
writer.add_scalar('psrn_y', psrn_y, update_step)
writer.add_scalar('ssim_rgb', ssim_rgb, update_step)
writer.add_scalar('ssim_y', ssim_y, update_step)
query = '''
INSERT INTO odesr01_04_val
(set14_psnr_rgb, set14_psnr_y, set14_ssim_rgb, set14_ssim_y,
set5_psnr_rgb, set5_psnr_y, set5_ssim_rgb, set5_ssim_y)
VALUES (%f, %f, %f, %f, %f, %f, %f, %f)
''' % tuple(metrics_list)
engine.execute(query)
model.train()
if not MODEL == 'BICUBIC':
saver = tf.train.Saver()
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
sess = tf.Session(config=config)
sess.run(tf.global_variables_initializer())
if not MODEL == 'BICUBIC':
saver.restore(sess, MODEL_CKPT_PATH)
fs = glob.glob(os.path.join(TEST_DIR, '*.bmp'))
psnrs = []
for f in fs:
img = misc.imread(f)
lr_img = misc.imresize(img, 1.0 / SCALE, 'bicubic')
lr_y = utils.rgb2ycbcr(lr_img)[:, :, :1]
lr_y = np.expand_dims(lr_y, 0).astype(np.float32) / 255.0
start = time.clock()
res_y = sess.run(res, feed_dict={lr: lr_y})
end = time.clock()
res_y = np.clip(res_y, 0, 1)[0] * 255.0
bic_img = misc.imresize(lr_img, SCALE / 1.0, 'bicubic')
bic_ycbcr = utils.rgb2ycbcr(bic_img)
bic_ycbcr[:, :, :1] = res_y
res_img = utils.img_to_uint8(utils.ycbcr2rgb(bic_ycbcr))
img_name = f.split(os.sep)[-1]
misc.imsave(os.path.join(OUTPUT_DIR, img_name), res_img)
gt_y = utils.rgb2ycbcr(img)[:, :, :1]
psnr = utils.psnr(res_y[SCALE:-SCALE, SCALE:-SCALE], gt_y[SCALE:-SCALE,
