To suppress stripe noise in selective overflow lateral overflow integration capacitor (SO-LOFIC) imaging, a spatially adaptive destriping method is proposed. The signal and noise characteristics of SO-LOFIC images are analyzed, and a degradation model is established to characterize column offset noise and region-dependent photoresponse nonuniformity (PRNU). An intensity-based weighting is introduced to distinguish different regions and enhance stripe extraction in high-intensity regions, while a structure-aware weighting derived from local gradient information avoids over-smoothing near edges. These weightings are integrated into a variational destriping model with column-group sparsity and direction-al smoothness constraints, and solved by alternating direction method of multipliers (ADMM). Experimental results on simulated data show average improvements of 0.8 dB in PSNR and 0.012 in SSIM over the second-best results across the dataset. Results on real data demonstrate effective stripe removal with high visual quality and detail preservation.