Recovery of Images with Missing Pixels using a Gradient Compressive Sensing Algorithm
This is an incremental improvement for image processing applications, addressing noise and missing pixel recovery in images.
The paper tackles image reconstruction from missing pixels by exploiting sparsity in the DCT domain, using a gradient-based compressive sensing algorithm to recover images affected by salt-and-pepper noise, with results compared to median filtering.
This paper investigates the possibility of reconstruction of images considering that they are sparse in the DCT transformation domain. Two approaches are considered. One when the image is pre-processed in the DCT domain, using 8x8 blocks. The image is made sparse by setting the smallest DCT coefficients to zero. In the other case the original image is considered without pre-processing, assuming the sparsity as intrinsic property of the analyzed image. A gradient based algorithm is used to recover a large number of missing pixels in the image. The case of a salt-and-paper noise affecting a large number of pixels is easily reduced to the case of missing pixels and considered within the same framework. The reconstruction of images affected with salt-and-paper impulsive is compared with the images filtered using a median filter. The same algorithm can be used considering transformation of the whole image. Reconstructions of black and white and colour images are considered.