Digital image correlation is a full-field, high-precision, non-contact optical technique that has been widely employed for full-field displacement and strain measurements under a broad range of testing conditions [[1], [2], [3], [4], [5], [6]]. By tracking the evolution of characteristic subsets on the specimen surface across a sequence of images, DIC enables accurate reconstruction of displacement and strain fields [7,8]. The deformation information carrier on the specimen surface may consist of either the natural texture of the test object or artificially fabricated speckle patterns [9,10]. However, features provided by natural textures are often limited by insufficient contrast and randomness, which can significantly degrade the accuracy of DIC computations. In contrast, artificially fabricated speckle patterns can be deliberately designed to overcome these limitations. Dong et al. [11] reviewed several mainstream speckle fabrication techniques, including spraying, spin coating, photolithography, focused ion beam processing, scratching, and polishing. These approaches inevitably cause a certain degree of damage to the material surface, which is difficult to restore and may introduce residual stresses, thereby affecting the measurement of physical quantities. Moreover, in certain specialised application scenarios, such as high-temperature environments, most coating-based artificial speckle patterns are prone to oxidation and fading. Under testing conditions involving large mechanical loads, mismatches in mechanical properties between the artificial speckle layer and the substrate may lead to speckle cracking or delamination. These factors not only degrade speckle image quality and compromise the accuracy of DIC measurements, but may even lead to the failure of DIC measurements altogether.

Laser speckle is a random intensity pattern formed on the CCD sensor when coherent light illuminates an optically rough surface and the reflected light is imaged through a lens. It is also referred to as subjective speckle and is directly associated with the displacement of the object surface [12]. Compared with conventional artificial speckle patterns, laser speckle does not suffer from fading or detachment under external thermal or mechanical loading, nor does it cause damage to the surface of the test specimen. Provided that the specimen surface is optically rough, laser speckle can serve as an effective alternative to artificial speckle patterns as the deformation carrier for DIC measurements. Yamaguchi [13,14] pioneered the use of laser speckle for measuring surface displacement and deformation of test specimens. Since then, laser speckle has been successfully employed as a substitute for artificial speckle patterns in a wide range of applications, including uniaxial tensile testing [15], vibration analysis [16], shear deformation measurements [17], and high-temperature thermo-mechanical deformation studies [18].

However, owing to the formation mechanism of laser speckle, pronounced local variations arise when the object surface undergoes relatively large deformation. As a result, the correlation between corresponding subsets in successive images deteriorates rapidly, leading to DIC matching failure, commonly referred to as the decorrelation effect. It is worth noting that decorrelation can also occur in large-deformation measurements when conventional artificial speckle patterns are used; nevertheless, laser speckle tends to decorrelate under much smaller loading levels. Previous studies [19] have reported that, in room-temperature uniaxial tensile tests employing laser speckle, decorrelation may occur when the specimen strain is as low as 2000 με (elongation corresponding to 0.002 times the original length), which falls well short of the measurement requirements in many practical applications.

To alleviate decorrelation, Anwander [20] proposed an incremental computation scheme in which correlation analyses are performed between successive images and the resulting strains are accumulated to determine the actual strain at each loading stage, successfully enabling the measurement of plastic deformation in aluminium specimens at elevated temperatures. However, this approach is limited to the evaluation of average strain and is incapable of providing full-field strain information. Pan [21] introduced an incremental accumulation scheme with reference image updating, in which the correlation coefficient between seed points in consecutive images is employed as the criterion for reference image updating. When combined with reliability-guided digital image correlation (RG-DIC), this method enabled successful full-field deformation measurement of foam specimens under compression. To reduce bias induced by reference image interpolation, Zhou [22] proposed a nearest-integer-pixel subset translation strategy, in which incremental computations are no longer performed using subsets centred at sub-pixel locations. Instead, the integer-pixel position closest to the sub-pixel location is selected as the new subset centre, and the resulting offset is compensated for during incremental calculations. This strategy eliminates grayscale interpolation during reference image updating and consequently reduces errors in incremental computations. Building upon this approach, Zhang [23] further improved measurement accuracy by introducing Gaussian pre-filtering. However, this filtering strategy inevitably degrades the spatial resolution of DIC measurements, and its reference image updating criterion based on manually selected seed points is poorly suited to handling local decorrelation induced by complex full-field strain distributions. Moreover, although this integer-pixel subset translation–based incremental scheme effectively reduces grayscale interpolation bias in the reference image, interpolation bias in the deformed images remains unavoidable.

To address these issues, this study proposes a high-accuracy incremental computation scheme. First, an adaptive reference image updating strategy based on the interquartile range (IQR) is introduced to overcome the limitation of conventional updating criteria that neglect local decorrelation, thereby reducing measurement distortion under complex strain conditions. In addition, an adaptive reference subset translation strategy is integrated with an interpolation bias compensation scheme to minimise the interpolation-induced bias introduced during each matching operation. This paper first presents the measurement principles of laser speckle DIC and the associated decorrelation effect. The proposed incremental computation scheme is then described in detail. Finally, the effectiveness of the proposed method is validated through simulated speckle experiments, in-plane translation tests, and uniaxial tensile experiments.