Abstract: Laser speckle is a new type of nondestructive detection technology, which can be used to characterize the activity of biological samples. Based on the laser speckle technique, an experimental platform was designed to investigate the feasibility of detecting the freshness of chilled pork. The optical maser with the wavelength of 465 and 660 nm and the power of 6 mw was used in this study. 84 pieces of chilled pork 24 h after slaughter, were cut into 6 cm × 6 cm × 8 mm and placed in a transparent dish with a lid. After numbering them, the samples were kept at 4 °C until measurement. For each sample, 640 laser speckle images with the resolution of 640 × 480 pixels were obtained at a regular interval of 0.06 s. The step was repeated every day for 7 days. Afterwards the laser speckle images were processed with a traditional method. Firstly the grey-scale processing was conducted, the fixed line in each gray images obtained was extracted, and the THSP (time history of speckle pattern) image was constructed using them according to the storage time. Then the co-occurrence matrix (COM) was calculated. Finally, the inertia moment (IM) of the COM was calculated and considered as the activity value of samples. The results showed that the IM values obtained by the traditional method frequently had abnormal values, and the computational stability of the method was rather poor. The IM values could be influenced by the traditional algorithm and the selection of different row, through which it was found that the traditional method was prone to outliers in the calculation of the correction matrix. Besides, the choices of different rows had great influence on IM values. The IM values presented Gaussian-like distribution from the center of the laser speckle to the around. And the IM values of odd or even lines under different excitation wavelengths were various due to the Bayer arrangement of CCD (charge coupled device) camera pixel. In this study, 3 kinds of improvements about traditional algorithms were proposed according to above findings. First, the sort algorithm was designed to calculate the IM values by dynamically selecting 3 lines with the highest activity in speckle images. Second, the calculation method of modified matrix for COM was improved. Finally, the method of calculating the distance between non-zero element and the diagonal was improved. The results exhibited that the outlier disturbance was effectively suppressed by the improved method. The speckle activity of chilled pork showed an upward trend firstly and then downward along with the shelf life. Besides, the activity changes of pork were more sensitive under laser wavelength of 660 nm than that under laser wavelength of 465 nm. According to the measured total volatile basic nitrogen (TVB-N) and speckle activity IM to establish pork freshness prediction model, and the results showed that the discrimination under the laser wavelength of 465 nm was better than the 660 nm laser, and the recognition rate for the training set and the prediction set could reach 87.50% and 89.29%, respectively. Therefore, using laser speckle technique to detect the freshness of cold pork is feasible, and the traditional method can be replaced by the improved method.