Adaptive deep residual network for single image super-resolution

In recent years,deep learning has achieved great success in the field of image processing.In the single image super-resolution(SISR)task,the convolutional neural network(CNN)extracts the features of the image through deeper layers,and has achieved impressive results.In this paper,we propose a single image super-resolution model based on Adaptive Deep Residual named as ADR-SR,which uses the Input Output Same Size(IOSS)structure,and releases the dependence of upsampling layers compared with the existing SR methods.Specifically,the key element of our model is the Adaptive Residual Block(ARB),which replaces the commonly used constant factor with an adaptive residual factor.The experiments prove the effectiveness of our ADR-SR model,which can not only reconstruct images with better visual effects,but also get better objective performances.

Single Image Dehazing with V-transform and Dark Channel Prior

Single image dehazing algorithm based on the dark channel prior may cause block effect and color distortion.To improve these limitations,this paper proposes a single image dehazing algorithm based on the V-transform and the dark channel prior,in which a hazy RGB image is converted into the HSI color space,and each component H,I and S is processed separately.The hue component H remains unchanged,the saturation component S is stretched after being denoised by a median filter.In the procession of intensity component,a quad-tree algorithm is presented to estimate the atmospheric light,the dark channel prior and the V-transform are used to estimate the transmission map.To reduce the computational complexity,the intensity component I is decomposed by the V-transformfirst,coarse transmission map is then estimated by applying the dark channel prior on the low frequency reconstruction image,and the guided filter is finally employed to refine the coarse transmission map.For images with sky regions,the haze removal effectiveness can be greatly improved by just increasing the minimum value of the transmission map.The proposed algorithm has low time complexity and performs well on a wide variety of images.The recovered images have more nature color and less color distortion compared with some state-of-the-art methods.