Image Sharpness Metric Based on Algebraic Multi-grid Method

Qian Ying,Ren Xue-Mei,Huang Ying,Meng Li

doi:10.14569/ijacsa.2014.050425

Abstract

In order to improve Mean Square Error of its reliance on reference images when evaluating image sharpness, the no-reference metric based on algebraic multi-grid is proposed. The proposed metric first reconstructs the original image by Algebraic Multi-grid (AMG), then compute the Mean Square Error between original image and reconstructed image, the result represents image sharpness. Experiments show that the proposed sharpness metric has better practicability and monotonicity, correlates well with the perceived sharpness. The algorithm has superiority in image sharpness metric.

Highlights

There has been an increasing need to develop quality measurement techniques that can predict perceived image/video quality automatically
We propose an improved Mean square error (MSE) together with reconstruction image use algebraic multi-grid
When applied Algebraic multi-grid in image reconstruction, first, we should convert the image into graph, create relationship affinity matrix on similarity between pixels gray value of image

Summary

INTRODUCTION

There has been an increasing need to develop quality measurement techniques that can predict perceived image/video quality automatically. These methods are useful in various image/video processing applications [1,2,3,4,5], such as compression, communication, printing, display, analysis, registration, restoration and enhance [6]. Ferzli[8] put forward perceptual sharpness metric based on measured just-noticeable blurs (JNBs), but unable to keep balance between stability and sensitivity. We propose an improved MSE together with reconstruction image use algebraic multi-grid. The clearer the image is, the smaller the similarity between pixels and the smaller of MSE between image reconstructed by algebraic multi-grid and Original Image. The metric proposed could used to measure image sharpness

PROPOSED NO-REFERENCE OBJECTIVE SHARPNESS METRIC

No-reference image sharpness metric based on AMG

RESULTS AND ANALYSIS

Prediction Accuracy

CONCLUSION