Nonlinear Kernel Based Feature Maps for Blur-Sensitive Unsharp Masking of JPEG Images
Jhilik Bhattacharya, TIET Patiala, India, jhilik@thapar.edu
Stefano Marsi, University of Trieste, Italy, marsi@units.it
Giovanni Ramponi, University of Trieste, Italy, ramponi@units.it
We propose a method for estimating the blur regions of an image, resorting to a mixture of linear and nonlinear convolutional kernels.
The blur map we obtain is utilized to enhance images such that the enhancement strength is an inverse function of the amount of measured blur.
The blur map can also be used for tasks such as attention-based object classification, low light image enhancement, and more.
We train a CNN architecture with nonlinear upsampling layers using a standard blur detection benchmark dataset, with the help of blur target maps.
Further, we propose to use the same architecture to build maps of areas affected by the typical JPEG artifacts, ringing and blockiness.
The blur map and the artifact map pair permit to build an activation map for the enhancement of a (possibly JPEG compressed) image.
Extensive experiments on standard test images verify the quality of the maps obtained using our algorithm and their effectiveness in locally controlling the enhancement, for superior perceptual quality.
Last but not least, computation time for generating these maps is much lower than the one of other comparable algorithms.