A Statistical Definition of Image Resolution Based on the Correlation of Pixels

TL;DR

This paper introduces a new statistical definition of image resolution based on pixel correlation, accounting for noise, object structure, and deconvolution effects, providing a more comprehensive measure than traditional criteria.

Contribution

The paper proposes a novel resolution definition that incorporates signal-to-noise ratio and object structure, and demonstrates its application in deconvolution analysis.

Findings

Weak structures can be extracted at low SNR with low resolution.

High SNR data yields high-resolution images after Wiener deconvolution.

The new definition aids in comparing deconvolution algorithms' effects.

Abstract

Resolution, usually defined by the Rayleigh criterion or the Full Width at Half Maximum of a Point Spread Function, is a basic property of an image. Here, we present a new statistical definition of image resolution based on the cross-correlation properties of the pixels in an image. It is shown that the new definition of image resolution depends not only on the PSF of an imaging device, but also on the signal-to-noise ratio of the data and on the structures of an object. In an image, the resolution does not have to be uniform. Our new definition is also suitable for the interpretation of the result of a deconvolution. We illustrate this, in this paper, with a Wiener deconvolution. It is found that weak structures can be extracted from low signal-to-noise ratio data, but with low resolution; a high-resolution image was obtained from high signal-to-noise ratio data after a Wiener deconvolution. The new definition can also be used to compare various deconvolution algorithms on their processing effects, such as resolution, sensitivity and sidelobe level, etc.