Estimating individual causal effects from observational data presents significant challenges, including data imbalance and confounding factors—particularly the distributional disparities between groups caused by intervention selection bias. Existing methods primarily aim to achieve global distributional balance, often overlooking the importance of preserving similarity information among individuals. In this paper, we propose a novel method for individual treatment effect estimation, termed Data-balance and Feature-extraction for Individual Treatment Effect (DFITE). The DFITE approach integrates data balance regularization constraints to map raw data into a balanced representation space, while simultaneously employing feature reconstruction to project high-dimensional data into a compact, low-dimensional space. This dual strategy enables both effective distributional alignment and informative feature extraction. Experimental results on the IHDP and Jobs datasets demonstrate that DFITE significantly improves the accuracy, stability, and generalizability of individual causal effect estimation, outperforming existing baseline methods. In summary, DFITE effectively addresses both distributional imbalance between intervention groups and the challenges of high-dimensional feature representation, providing more accurate and robust estimates of individual treatment effects.