Perceptual Variance Weight Matrix based Adaptive Block Compressed Sensing for Marine Image Compression

The underwater marine environment is made up of a huge number of interconnected, resource-limited underwater equipment capable of monitoring enormous, unknown water bodies. These devices, in particular, are outfitted with cameras to capture underwater landscapes and interact with one another. However, the amount of data created is enormous, limiting the devices' computational capability and battery life. To unravel the issues, extreme high compression is required. Adaptive block compressed sensing (ABCS) is a subcategory of compressed sensing (CS) in which sampling and compression is performed at sub-nyquist rate. ABCS can achieve better compression and sampling performance than CS. To render high quality to the reconstructed image components, variance between the image pixels are utilized to construct the perceptual weight matrix. This perceptual variance weight matrix is applied on the image vector to select image components which attracts human eye. To achieve high quality reconstruction and better compression, combination of ABCS and perceptual variance weight matrix (PWM-ABCS) is proposed in this paper.

Automated summary

This paper introduces the device issues and data compression techniques in the underwater marine environment. By combining adaptive block compressed sensing with the perceptual variance weight matrix, high-quality image reconstruction and better data compression performance are achieved.