This paper proposes two watermarking-based algorithms for temper detection and recovery of the tampered regions, which can be applied to official documents, such as digitized passports and governmental registrations. In both algorithms, a halftone version of the original grayscale image is used as an approximated version of the host image (image digest), which is then embedded as a watermark sequence into the frequency domains of the host image. In the first algorithm, the Integer Wavelet Transform (IWT) is used for watermark embedding (WIA-IWT); while in the second one, the Discrete Cosine Transform (DCT) domain is used for this purpose (WIA-DCT). The experimental results demonstrate the robustness of both algorithms against content preserved modifications, such as JPEG compression, as well as recovery capability. Moreover, we employ a Multilayer Perceptron neural network (MLP) in inverse halftoning process to improve the recovered image quality, where the gray-scale of the modified area is estimated by the MLP using the extracted halftone image. The experimental results demonstrate the effectiveness of the proposed scheme.