Fully Point-wise Convolutional Neural Network for Modeling Statistical Regularities in Natural Images
This work addresses a domain-specific problem in image processing by providing a more efficient CNN architecture for modeling statistical regularities, though it is incremental as it builds on existing deep learning approaches.
The paper tackles the problem of inaccurate statistical representation in CNNs due to pixel intensity variability by introducing a fully point-wise CNN architecture that uses shuffled images as input. The result is a network that achieves comparable performance on color constancy and image dehazing tasks while using 1/10 to 1/100 of the parameters and computational cost of existing methods.
Modeling statistical regularity plays an essential role in ill-posed image processing problems. Recently, deep learning based methods have been presented to implicitly learn statistical representation of pixel distributions in natural images and leverage it as a constraint to facilitate subsequent tasks, such as color constancy and image dehazing. However, the existing CNN architecture is prone to variability and diversity of pixel intensity within and between local regions, which may result in inaccurate statistical representation. To address this problem, this paper presents a novel fully point-wise CNN architecture for modeling statistical regularities in natural images. Specifically, we propose to randomly shuffle the pixels in the origin images and leverage the shuffled image as input to make CNN more concerned with the statistical properties. Moreover, since the pixels in the shuffled image are independent identically distributed, we can replace all the large convolution kernels in CNN with point-wise ($1*1$) convolution kernels while maintaining the representation ability. Experimental results on two applications: color constancy and image dehazing, demonstrate the superiority of our proposed network over the existing architectures, i.e., using 1/10$\sim$1/100 network parameters and computational cost while achieving comparable performance.