Edge preserving image smoothing benchmark system with deep convolutional neural network

A single Edge preserving image smoothing cannot perform well for a wide range of image datasets. The performance evaluation of edge preserving image smoothing algorithm varies based on the parameter chosen and the image data set. This invention discloses Edge Preserving Image Smoothing Benchmark System with Deep Convolutional Neural Network comprising of Training Phase (303) and Testing Phase (304). The Training Phase (303) comprising of Labels (401), Training Image Set (402) Smoothing Filter (403), Feature Extraction (404), VDCNN-ResNet (405), and Loss Function (406). The Testing Phase (304) comprising of Test Image (501), VDCNN ResNet (502), and Filtered Image (503). The Edge Preserving Image Smoothing Benchmark System with Deep Convolutional Neural Network uses the image dataset with ground truth image smoothing results and Baseline Algorithms for generating Edge Preserving image smoothing. The Collection of Smoothing Algorithms representing various categories are linear filtering, non-linear filtering, partial differential equation filtering, hybrid filtering, filtering methods based on DWT, and global optimization based filtering can produce high quality smoothing over wide range of inputs. The established dataset contains 500 training and test images while the baseline methods in our invention are built on representative Deep Convolutional Neural Network architecture with VDCNN and ResNet.