Restoration of Simple Lens Captured Images based on Deep Multi-Scale Network

Inspired by computational imaging technology, simple lens can be combined with the image restoration algorithm to achieve satisfactoy imaging quality. Therefore, the conventional complex lens can be replaced with simple lens to reduce the cost and weight. However, images captured by simple-lens cameras suffer from spatial varying blur. An image restoration algorithm is proposed based on a deep multi-scale network for simple lens imaging systems. Image feature are processed and analyzed at multiple resolution levels to improve the learning ability of the network. The experimental and results prove that the image restoration algorithm we proposed can achieve promising restoration results.