Review: CNN for JPEG Compressed Image Deblocking (Codec Post-Processing)

With Deep CNN, Outperforms ARCNN

Sik-Ho Tsang
Aug 2 · 3 min read

In this story, a deep convolutional neural network (CNN) for compressed image deblocking, by Fudan University, is briefly reviewed. With the use of CNN, the decoded/reconstructed image can obtain higher PSNR. This is the paper in 2017 ICME. (Sik-Ho Tsang @ Medium)


  1. Network Architecture
  2. Experimental Results

1. Network Architecture

Network Architecture
  • The proposed model consists of 20 convolutional layers with taking Rectified Linear Unit as activation function. Each convolutional layer has 32 feature maps and the size of the convolution kernel is 5 × 5.
  • Short skip connections and a long skip connection are used as shown above, inspired by ResNet.

After Decoding the JPEG as raw image, the decoded image goes through the proposed network for deblocking to obtain better image quality.

  • The proposed network converges after about 2 × 10⁶ backprops.
  • However, we can see that AR-CNN and SRCNN both converge after about 4.5 × 10⁸ backprops.

2. Experimental Results

  • The proposed network obtains the highest PSNR, SSIM and PSNR-B.
  • It outperforms ARCNN should be due to the reason of using a deeper network compared with ARCNN.
Some Visualizations: Ground Truth (Leftmost), JPEG (Second Left), ARCNN (Second Right), Proposed Network (Rightmost)
  • The proposed network can have sharper (less blurry) images.


[2017 ICME] [Li ICME’17]
An Efficient Deep Convolutional Neural Networks Model for Compressed Image Deblocking

My Previous Reviews

Image Classification [LeNet] [AlexNet] [Maxout] [NIN] [ZFNet] [VGGNet] [Highway] [SPPNet] [PReLU-Net] [STN] [DeepImage] [SqueezeNet] [GoogLeNet / Inception-v1] [BN-Inception / Inception-v2] [Inception-v3] [Inception-v4] [Xception] [MobileNetV1] [ResNet] [Pre-Activation ResNet] [RiR] [RoR] [Stochastic Depth] [WRN] [Shake-Shake] [FractalNet] [Trimps-Soushen] [PolyNet] [ResNeXt] [DenseNet] [PyramidNet] [DRN] [DPN] [Residual Attention Network] [DMRNet / DFN-MR] [IGCNet / IGCV1] [MSDNet] [ShuffleNet V1] [SENet] [NASNet] [MobileNetV2]

Object Detection [OverFeat] [R-CNN] [Fast R-CNN] [Faster R-CNN] [MR-CNN & S-CNN] [DeepID-Net] [CRAFT] [R-FCN] [ION] [MultiPathNet] [NoC] [Hikvision] [GBD-Net / GBD-v1 & GBD-v2] [G-RMI] [TDM] [SSD] [DSSD] [YOLOv1] [YOLOv2 / YOLO9000] [YOLOv3] [FPN] [RetinaNet] [DCN]

Semantic Segmentation [FCN] [DeconvNet] [DeepLabv1 & DeepLabv2] [CRF-RNN] [SegNet] [ParseNet] [DilatedNet] [DRN] [RefineNet] [GCN] [PSPNet] [DeepLabv3] [LC] [FC-DenseNet] [IDW-CNN] [SDN]

Biomedical Image Segmentation [CUMedVision1] [CUMedVision2 / DCAN] [U-Net] [CFS-FCN] [U-Net+ResNet] [MultiChannel] [V-Net] [3D U-Net] [M²FCN] [SA] [QSA+QNT] [3D U-Net+ResNet]

Instance Segmentation [SDS] [Hypercolumn] [DeepMask] [SharpMask] [MultiPathNet] [MNC] [InstanceFCN] [FCIS]

Super Resolution [SRCNN] [FSRCNN] [VDSR] [ESPCN] [RED-Net] [DRCN] [DRRN] [LapSRN & MS-LapSRN] [SRDenseNet]

Human Pose Estimation [DeepPose] [Tompson NIPS’14] [Tompson CVPR’15] [CPM]

Codec Post-Processing [ARCNN] [Li ICME’17]

Sik-Ho Tsang

Written by

PhD, Researcher. I share what I've learnt and done. :) My LinkedIn:

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade